,QWHUQDWLRQDO6\PSRVLXPRQ&RPSXWHU

6FLHQFH&RPSXWHU(QJLQHHULQJDQG
(GXFDWLRQDO7HFKQRORJ\

Prof. Dr. Wolfram Hardt

TUDpress
IBS Scientific Workshop Proceedings
Herausgegeben von Stiftung IBS, Wolfram Hardt
Band 
,QWHUQDWLRQDO6\PSRVLXPRQ
&RPSXWHU6FLHQFH&RPSXWHU
(QJLQHHULQJDQG(GXFDWLRQDO
7HFKQRORJ\
Prof. Dr. Wolfram Hardt (Hrsg.)

TUDpress
2020
%LEOLRJUDȴVFKH,QIRUPDWLRQGHU'HXWVFKHQ1DWLRQDOELEOLRWKHN
'LH'HXWVFKH1DWLRQDOELEOLRWKHNYHU]HLFKQHWGLHVH3XEOLNDWLRQLQGHU
'HXWVFKHQ1DWLRQDOELEOLRJUDȴHGHWDLOOLHUWHELEOLRJUDȴVFKH'DWHQVLQGLP
,QWHUQHW¾EHUKWWSGQEGQEGHDEUXIEDU

%LEOLRJUDSKLFLQIRUPDWLRQSXEOLVKHGE\WKH'HXWVFKH1DWLRQDOELEOLRWKHN
7KH'HXWVFKH1DWLRQDOELEOLRWKHNOLVWVWKLVSXEOLFDWLRQLQWKH'HXWVFKH1D
WLRQDOELEOLRJUDȴHGHWDLOHGELEOLRJUDSKLFGDWDDUHDYDLODEOHLQWKH,QWHUQHW
DWKWWSGQEGQEGH

,6%1

k78'SUHVV
7+(/(08QLYHUVLW¦WVYHUODJ
*PE+XQG&R.*
KWWSZZZWXGSUHVVGH

$OOH5HFKWHYRUEHKDOWHQ$OOULJKWVUHVHUYHG
*HVHW]WYRQGHQ+HUDXVJHEHUQ
3ULQWHGLQ*HUPDQ\
IBS Scientific Workshop Proceedings
The “International Symposium on Computer Science, Computer Engineering and
Educational Technologies” (ISCSET-2020) is the latest edition of the annual
international symposium on computer science funded by the “Internationales
Informatik- und Begegnungszentrum Sachsen (IBS)” fund and organized since 2012.
It was our distinct pleasure for this year’s symposium to organize by the Mongolian
Institute for Educational Research of the Ministry of Education and Science in
cooperation with IBS.
The successful organization of ISCSET-2020 allowed us to provide an opportunity for
leading international researchers to exchange ideas and research outcomes, facilitating
knowledge exchange among the research community.
While this year’s symposium took place at a time of a new normal brought by the
COVID-19 pandemic that has ravaged the world, it brought together leading
researchers in the fields of Computer Science and Educational Technologies in order
to discuss theories, practices, exchange ideas and introduce outcomes of innovative
research.
It is my strong belief that the issues discussed in this symposium will not only allow
us to overcome unprecedented challenges but in turn will make a contribution to
opening the door for citizens of the world to access high quality education in their field
of interest from anywhere in the world.
Even though the symposium took place in an online manner this year due to the
COVID-19 pandemic, we would like to extend our appreciation to the researchers that
took part in the symposium and presented innovative ideas. From the bottom of our
hearts, we would like to wish you endless success in your future works.
We will always be eager to further cooperate with you.

Dr. Lkhagvasuren Purev

Director, Mongolian Institute for Educational Research
Prof. Dr. Dr. h. c. Uranchimeg Tudevdagva
Mongolian University of Science and Technology,
Chemnitz University of Technology
October 2020
Table of Contents

International Symposium on Computer Science, Computer Engineering and

Educational Technology

Deep Learning for Personal Protective Equipment Detection ………..……………………… 9

Artem Soroka, German Magay, Bukhamer Egor

Incorporating Continuity between Professional Subjects into the English Curriculum ……... 14
Batbaatar Dashdemberel

Education in Pandemic Crisis Implementation via ICT Online Technologies ………………. 21

Cyryl Koshyk

Risks in the Educational Sector: Theory and Practice ……………………………………….. 25

Enkhtuvshin Jambaldorj

Sample Model for Selecting State Examination Test Item …………………………………... 29

Gan-Erdene Jigjid, Ganbat Tsend

Perception of Lane Curvature Detection for Situational Awareness ………………………… 32

Julkar Nine, Shanmugapriyan Manoharan, Mohammad Didar Hossain Sagar,
Shadi Saleh, Wolfram Hardt

MRI Rendering with EEG Sensors Projection ……………………………………………….. 36

Klementev Dmitrii

Analyses on Physics Textbooks for Secondary Schools ……………………………………… 40

Munkhbold Mangalsuren, Purevdorj Dondovsambuu, Sodov Tsedendamba

Autonomous Unmanned Aerial Vehicle Development: MAVLink Abstraction Layer ………. 45

Marco Stephan, Batbayar Battseren, Uranchimeg Tudevdagva

Comprehensive Professional Competence Study of Legal University Professors ……...……. 50

Narantuya Lkhundev, Sodov Tsedendamba

Assessing the Progress of Children’s Competencies …………………………...…………….. 53

Narantuya Jamyandorj, Ganbat Tsend

Analysis of User Interface Design Methods ………………………………………………….. 57

Nikolai Gervas

Developing an Information Culture for Teachers ……………………………………………. 61

Nyamsuren Аmsarvaa, Tsedendamba Sodov
“Mathshop” E-Learning Mathematics Program - Experiments and Opportunities for Future
Development …………………………………………………………………………………. 65
Otgontugs Bayarrintsen, Erdenebat Dashtseren, Lkhagvasuren Purev, Bilguunbaatar Tumurkhuyag

Collision Warning Based on Multi-Object Detection and Distance Estimation ……………… 68

Shadi Saleh, Chirag Rellan, Shubham Pareshkumar Surana, Julkar Nine, Wolfram Hardt

The Study to Determine the Required Product Identification System ……………………….. 73

Sod-Od Batzorig, Gundsambuu Bold

Comparison of Photogrammetric Approaches for Estimating Motion Parameters

from a Monocular Sequence …..……………………………………………………………… 77
Tserennadmid Tumurbaatar

Studying the Students Information Technology Use ………………………………………… 81

Tuvshinjargal Radnaa, Tserengavaa Tserenpurev

Blended Mentoring in Academic Internship Program ……………………………………….. 84

Ummay Ubaida Shegupta, René Schmidt, Martin Springwald, Wolfram Hardt

The Evaluation of Distance Teaching During COVID Quarantine at MUST ………………… 90

Uranchimeg Tudevdagva, Tsooj Shambaljamts, Delgermaa Senden
 Deep Learning for Personal Protective Equipment
Detection
Artem Soroka German Magay Bukhamer Egor
Novosibirsk State Technical University Novosibirsk State Technical University Novosibirsk State Technical University
Novosibirsk, Russia Novosibirsk, Russia Novosibirsk, Russia
[email protected] [email protected] [email protected]

Abstract—To prevent the consequences of emergencies at one-stage approach were used to implement the detection and
the workplace, it is necessary to use personal protective recognition of personal protective equipment instances in
equipment (PPE). There is an important problem of images.
automating personal protective equipment detection. In this
A.MobilenetV3
paper, the problem of automating PPE detection is solved
by using the modern architecture of RetinaNet to detect and The mobilenetV3 model is a representative of a new
classify objects in the image. generation of neural networks, which is based on a modern
architecture designed for low computing resources. Searching
Keywords—deep learning; detecting; classification; for MobileNetV3 [7] describes implementations of the latest
safety; loss function. version of mobile net: MobileNetV3-Large and
MobileNetV3-Small. The model is intended for semantic
I. INTRODUCTION image segmentation and object detection.
There are many hazards at workplaces that can harm the MobileNetV3, shows better results compared to
worker, such as high voltage, falling objects, sparks, etc. For MobileNetV2 [8]: on the MobileNetV3-Large modification it
example, wearing a hardhat is an effective protective measure is more accurate in the classification problem on the
for minimizing the risk of traumatic brain injury. Moreover, ImageNet dataset, and objects are detected more than 25%
a survey conducted by the US Bureau of Labor Statistics faster and with approximately the same accuracy as on
(BLS) suggests that 84% of workers who had suffered impact MobileNetV2, the experiment was carried out on the COCO
injuries to the head were not wearing head protection dataset [9].
equipment [1].
Compared to the previous implementation, MobileNetV3
To ensure the safety of employees at work, there is a need has a set of fundamental innovations: changing the structure
to use personal protective equipment (PPE) to prevent of the bottleneck layer, adding a new activation function, as
situations that threaten the health and life of the employee. As well as moving the expanding layer (bottleneck) beyond the
a result, there is an urgent task of automation of PPE merge level.
detection.
B.YOLOv3
As the field of deep learning and computer vision evolves,
more and more advanced image processing techniques are YOLO or «You Only Look Once» is CNN (convolutional
emerging. neural network) very popular architecture, which is used to
recognize multiple objects in an image. The main feature of
In this paper, the task of PPE detecting is solved by using this architecture compared to others is that most systems
modern neural network architectures to detect objects in the apply CNN several times to different regions of the image; in
image and classify instances. YOLO, CNN is applied once to the entire image at once. The
network divides the image into a kind of grid and predicts
II. OVERVIEW OF ADVANCED DEEP LEARNING SOLUTIONS bounding boxes and assigns it to certain object class.
When solving computer vision problems, two types of
approaches to detection (localization, object detection) and YOLOv3 is an improved version of the YOLO
recognition are used. One-stage approach: this approach uses architecture. It consists of 106 convolutional layers and better
one convolutional neural network to detect an object and detects small objects compared to its predecessor YOLOv2.
assign it to a class. These detectors include SSD [2] and The main feature of YOLOv3 is that there are three layers
YoloV3 [3]. at the output, each of which is designed to detect objects of
On the contrary, two-stage involves 2 stages of object different sizes.
detection. First, the model proposes a set of regions of С. RetinaNet
interests (ROI) by select search or regional proposal network. The RetinaNet [6] neural network differs from other
The proposed regions are sparse as the potential bounding algorithms in a set of fundamental features that allow it to
box candidates can be infinite. Second, a classifier only show good results and outperform competitors in terms of
processes the region candidates. Examples of architectures accuracy, the innovations proposed by RetinaNet authors will
based on this approach are Faster R-CNN [4] or Mask-RCNN be described below. Focal loss function with hyper parameter.
[5]. In the paper [6], the authors consider a modification of the
In this work, neural networks with the architecture loss function (1) for classification problems - Cross-entropy
RetinaNet [6], YoloV3 [3] and MobilenetV3 [7] based on the (CE).

Copyright ©2020 by IBS Foundation 9

 CE( ) = −log( ) (1)

The authors pay attention to the fact that in the classical

cross-entropy function the influence of wrong examples is
poorly controlled and propose to supplement the function
with a multiplier-regulator "penalty", the coefficients of
which are hyperparameters.

FL( ) = −(1 − ) log( ) (2)

γ – loss function hyperparameter, – probability of

ground truth class.
There is also a modification of Focal loss with an
additional balance coefficient α, within the limits of this article
the parameter α will be selected by experiment.

FL( ) = −α (1 − )γ log( ) (3)

In the article, the authors have chosen the best hyper Fig. 2. RetinaNet network architecture.
parameter value γ = 2. Fig. 1 shows the graph from work [6],
showing the dependence of losses on the accuracy of The specific feature of the pyramidal structure of the
prediction at different values of hyper parameter γ. As can be artificial neural network is that the features extracted by the
seen, the transition from the value of the loss indicator to a convolution layers of the map at different levels of the
high probability of belonging to the class, with the growth of convolution layers of the neural network are considered as a
hyperparameter γ becomes more complicated, i.e. high hierarchical pyramid. Maps at the lower levels have a high
probability of belonging to the class is achievable at resolution, and the upper levels have a high semantic
sufficiently low values . component. Maps of the upper level are up sampled to the
size of the lower layer map, and then combined into a single
set of maps, a similar operation is performed for all layers.
This algorithm structure provides sufficiently high quality
performance against the other modern detector architectures.
Table I shows a comparison table of neural networks and their
mAP (mean Average Precision) indicators at different IoU
(Intersection over Union) - the overlap between the areas of
prediction and validation example.

TABLE I. MAP VALUES FOR THE DIFFERENT ARCHITECTURES

Indicators
Methods Backbone
AP AP50 AP75
Two-stage methods
Fig. 1. Dependence of the probability of belonging to a class on the loss Faster R-CNN+++ ResNet-101-C4 34.9 55.7 37.4
function at different values of hyper parameter γ.
Faster R-CNN w FPN ResNet-101-FPN 36.2 59.1 39.0
Faster R-CNN by Inception-ResNet-v2 34.7 55.5 36.7
In our work further we will consider an experiment with G-RMI
the selection of hyper parameters and γ, the purpose of which Faster R-CNN w Inception-ResNet-v2- 36.8 57.7 39.2
is to achieve the best possible neural network performance on TDM TDM
its own data set. One-stage methods
YOLOv2 DarkNet-19 21.6 44.0 19.2
Moreover, RetinaNet architecture is based on a pyramidal SSD513 ResNet-101-SSD 31.2 50.4 33.3
structure. The backbone (basic model) for RetinaNet feature DSSD513 ResNet-101-DSSD 33.2 53.3 35.2
map extraction is FPN (Feature Pyramid Network) [10], a RetinaNet ResNet-101-FPN 39.1 59.1 42.3
modification of ResNet architecture [11]. RetinaNet ResNeXt-101-FPN 40.8 61.1 44.1
YOLOv3 608x 608[3] Darknet-53 33.0 57.9 34.4
Also RetinaNet includes two sub networks, the first sub
network performs a concurrent classification of objects at the
backbone outputs, the second, regression of the boundaries of In the article [3] the authors compare indicators of the
the detected object. The activation function of each approach used in RetinaNet and other modern neural
convolution layer is ReLu, the sigmoidal activation function networks with a one-stage YOLOv3 detector. From the Table
is set at the classifier output. Fig. 2 shows structural scheme I, we can see that RetinaNet shows the best results exactly,
of RetinaNet. but loses in frame processing speed. As for the task set in this

10
article, the focus is on object detection accuracy, sacrificing range [0:1]. mAP is calculated for a specific IoU, in this paper
speed. IoU will be taken more or equal to 0.5.

III. TRAINING RETINANET, MOBILENETV3, YOLOV3 F1=2×(Precision×Recall)/(Precision+Recall) (4)

When implementing a system of recognition of PPE, In this paper, due to the Focal Loss functions of the
training is provided on its own dataset: video from production RetinaNet neural network, we will also initialised
hyperparameters of loss functions, the coefficients αt and γ. To
and images of employees with and without PPE. On the
select the optimal value of the hyper parameters of the loss
images were marked areas of PPE, the detector is trained on
function, a search on lattice was performed. The domain of
13 classes: the presence / absence of PPE (positive and definition of hyper parameters was established based on the
negative examples), as well as the presence of a person. The best combinations of parameters in the article [6]: αt = 0.25, γ
data set was marked using the Visual Object Tagging Tool. = 2.0:
The dataset was randomly split into a test dataset and a α = {0.05,0.15,0.25,0,35}
learning dataset at a ratio of 800 examples – learning data,
200 - testing data. The total number of instances in all images γ={0.5,1.0,2.0,3.0,4.0}
is more than 4800 A search was performed for each pair (αt, γ). A series of
The trained neural network model makes a prediction of experiments was performed with the initialisation of different
finding a specific class of PPE in a certain zone. parameters αt and γ of the loss function in order to increase the
mAP50 values. Table II shows the dependence of the quality
The process of training of detectors requires certain indicators of neural network learning with initialised hyper
hyperparameters to be set in advance, such as the number of parameters, in Fig. 3 shows a graph of dependence. As can be
epochs, learning rate, the size of the batch, the optimization seen from the table, the best combination of parameters αt =
method. 0.15, γ = 2 gives AP50 = 58.92.
Number of epochs increasing leads to an increase in TABLE II. TABLE OF THE DEPENDENCE OF THE INDICATORS OF
accuracy, but large number of epochs significantly increases THE QUALITY OF LEARNING A NEURAL NETWORK WITH
INITIALISED HYPER PARAMETERS.
the training time without a noticeable improvement in the
quality of prediction. If the loss values decreased slightly over Hyperparameters Indicators
several epochs in a row training was stopped. The size of the αt γ AP50 Recall precision F-score
batch is the number of images at the input of the neural 0.05 2.0 48.60 0.520 0.394 0.449
network in one pass, was set equal to 2. All hyper parameters 0.15 1.0 53.92 0.510 0.554 0.531
were initialized on the basis of experiments in order to achieve
0.15 2.0 58.92 0.691 0.387 0.496
the optimal ratio between high quality training and the least
0.15 3.0 54.00 0.688 0.230 0.344
time spent.
0.25 0.5 51.25 0.495 0.515 0.505
To minimize the Focal loss quality functional, one of the 0.25 1.0 55.02 0.561 0.509 0.534
effective methods of stochastic optimization was used – Adam 0.25 2.0 58.44 0.706 0.424 0.530
[12], learning rate – 3×10-4, the model was trained in 30 0.25 3.0 55.14 0.748 0.233 0.355
epochs. 0.25 4.0 54.76 0.775 0.206 0.325
A. Hardware and Software 0.35 1.0 50.47 0.605 0.425 0.499
0.35 2.0 57.09 0.690 0.377 0.488
The personal protective equipment detection system was
0.35 3.0 54.78 0.773 0.214 0.335
implemented in the Python 3 programming language using the
Pytorch deep learning library, training visualization tools, the
TensorboardX package, plotly, OpenCV library for computer
vision and image processing algorithms and Nvidia CUDA
technology.
Often, training and testing deep learning models requires
significant computational resources. The use of graphic
process units, the advantage of which is parallel computing,
can reduce the training time, so the model was trained on the
Nvidia GeForce GTX 2060 GPU.
In addition to the GPU, the hardware characteristics of the
PC: Intel Core i7-7900f processor; RAM 16.0 GB; GPU
frequency 1710 MHz, 8192 MB GDDR6 video memory.
B. Experimental results for RetinaNet, MobilenetV3, YOLOv3
To assess the quality of the trained model, the following
indicators were chosen: recall, precision, the main criteria for
the quality of training are their generalization - F-measure (f1 Fig. 3. Graph of mAP versus hyper parameters of the loss function.
score) and Average precision (AP). Recall is an indicator of
the sensitivity of the algorithm to type II errors (false We trained the RetinaNet model at 100 epochs with the
negative). Precision is an indicator of the sensitivity of the best parameters and got the value mAP50 = 60.08, which is
learning algorithm to type I errors (false positives). F-measure comparable to the value mAP50 presented in the original work
can be interpreted as the integral assessment of accuracy and [6]. The models of neural networks YOLOv3 and
recal. mAP calculates the average precision for recall in the MobileNetV3 were also trained. The indicators of the quality

11
of neural networks learning is presented in Table III, the absence of hardhat on workers [13], a method for detecting
results of the learning algorithm are presented in Fig. 4, 5. the presence of hardhat using the SSD neural network
architecture [14], one of the most recent approaches based on
TABLE III. CNN METHODS COMPARISON. YOLOv3 [16]. Table 4 shows comparison between previous
Model Indicators papers and ours.
architecture mAP50 Recall Precision F1-score
TABLE IV. PREVIOUS STUDIES COMPARISON. THE
MobileNetV3 43.81 0.523 0.371 0.449 PERFORMANCE OF THE PROPOSED METHODS IN TERMS OF
YOLOv3 56.35 0.717 0.252 0.355 AVERAGE PRECISION (AP50) AND DETECTION SPEED.
RetinaNet 60.08 0.691 0.387 0.496
Criteria Hat Vest/Uniform
No No FPS
Detection Hat Vest
hat vest
previous studies
Fang et al. [13] >90,00p × × 10
Xie et al. [15]
54,60 × × 11
(CAHD)
Wu et al. [14]
74,76 62,67 × × 3,22
(Pelee-RPA)
Nath et al. [16]
74,29 63,84 74,32 73,18 13
(Approach-1)
Ours 79,90 76,67 80,48 79,98 7
× - Not applicable
p - Precision.

As seen at Table IV approach presented by us is better in

some criteria to the existing approaches. However, it should
be noted that our approach covers more categories. Besides,
it is necessary to note the knock-out values to determine the
position of the helmet visor. The supposed problem is its
transparency, for which it was decided to add few stickers to
the visor, which increased the AP value for determining the
position by several points. However, making changes to such
objects is prohibited by regulatory documentation, and in case
of making it requires a long-term approval, therefore, at this
stage of development, we are forced to discard that decision.
Table V shows the average precision value for each class for
Fig. 4. The result of the neural network our method.

TABLE V. THE PERFORMANCE OF OUR METHOD IN TERMS OF

AVERAGE PRECISION (AP50).

Detected object class AP50

helmet 79,90
no_helmet 76,67
gloves_on 66,31
gloves_removed 71,32
work_uniform 80,48
no_uniform 79,98
work_pants 78,31
no_work_pants 77,89
work_boots 29,41
no_work_boots 27,47
visor_rised 17,12 (42,78S)
visor_down 12,29(50,14S)
person 83,89
mAP (mean) 60,08
s – with stickers

The existing Nath et al. [16] YOLOv3 based approach is

a good example with high performance for fast work, but our
approach gives more accurate prediction, but sacrificing at
speed. This was more suitable for our task.
IV. POSSIBLE SOLUTION MODIFICATIONS
Fig. 5. The result of the neural network To assess the presence of PPE on a person in the frame, a
heuristic approach (kind of decision tree) was used, which can
Recently, methods based on the neural network approach be described as follows, for example, for a helmet:
for solving the problems of detecting PPE are gaining more As a result of the work of the neural network, we have an
and more popularity. A number of methods can be cited as understanding of the presence of a person in the image and its
examples. Based on Faster R-CNN architecture, to detect the relative coordinates, because the model is trained to localize

12
the image of a person too. At the same step, we know the [4] S. Ren, K. He, R. Girshick, J. Sun, Faster R-CNN: Towards
coordinates of the head with the helmet or its absence (in the Real-Time Object Detection with Region Proposal Networks,
training sample, there are classes - “with helmet”, “without in IEEE Transactions on Pattern Analysis and Machine
helmet”). Further, knowing the coordinates of the “person” Intelligence, vol. 39, no. 6, pp. 1137-1149, (2017).
object in the image, there is an understanding that the “helmet” [5] K. He, G. Gkioxari, P. Dollár, R. Girshick, Mask R-CNN,
object should not go beyond the boundaries of the “person” 2017 IEEE International Conference on Computer Vision
object. (ICCV), Venice, pp. 2980-2988, (2017).
[6] T. Lin, P. Goyal, R. Girshick, K. He, P. Dollár, Focal Loss for
In cases where it is necessary to expand the functions of Dense Object Detection, 2017 IEEE International Conference
the video analytics system, optionally, a variant of the on Computer Vision (ICCV), Venice, pp. 2999-3007, (2017).
composition of the detector algorithm and the algorithm for [7] A. Howard, M. Sandler, G. Chu, L.-C. Chen, B. Chen, M. Tan,
determining a person’s posture is possible, which allows you W. Wang, Y. Zhu, R. Pang, V. Vasudevan, Q. V. Le, H. Adam.
to analyze a person’s posture to assess and prevent accidents Searching for MobileNetV3 (cite arxiv: 1905.02244), (2019).
in hazardous situations: employee fall, immobility, unusual [8] M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, L. Chen,
behavior, etc. MobileNetV2: Inverted Residuals and Linear Bottlenecks,
2018 IEEE/CVF Conference on Computer Vision and Pattern
To determine the coordinates of body parts and the human Recognition, Salt Lake City, UT, 2018, pp. 4510-4520,
posture, the OpenPose algorithm [17] is used. The (2018).
implemented algorithm has the possibility of several people [9] T.-Y. Lin et al. (2014) Microsoft COCO: Common Objects in
being in the image. At the output of the convolutional neural Context. In: Fleet D., Pajdla T., Schiele B., Tuytelaars T. (eds)
network, we get a discrete set of reference points that Computer Vision – ECCV 2014. ECCV 2014. Lecture Notes
correspond to parts of the body, and then we analyze the in Computer Science, vol 8693. Springer, Cham, (2014).
associations of pairs of points. [10] T. Lin, P. Dollár, R. Girshick, K. He, B. Hariharan, S.
Belongie, Feature Pyramid Networks for Object Detection,
V. CONCLUSION 2017 IEEE Conference on Computer Vision and Pattern
Recognition (CVPR), Honolulu, HI, 2017, pp. 936-944,
Modern methods of data analysis and computer vision can
(2017).
be applied to effectively monitor compliance with safety [11] K. He, X. Zhang, S. Ren, J. Sun, Deep Residual Learning for
standards by employees, which will lead to a reduction in Image Recognition, 2016 IEEE Conference on Computer
emergency situations in hazardous industries. As part of this Vision and Pattern Recognition (CVPR), Las Vegas, NV,
work, an approach using deep learning tools to the task of 2016, pp. 770-778, (2016),
determining the availability of personal protective equipment [12] D. Kingma, J. Ba, Adam: A Method for Stochastic
for employees was proposed: a training data set was prepared, Optimization, International Conference on Learning
the neural network architecture was selected and the optimal Representations (2014).
combination of hyper parameters was established. The value [13] Q. Fang, H. Li, X. Luo, L. Ding, H. Luo, T.M. Rose, W. An,
Detecting non-hardhat-use by a deep learning method from
of the average precision of the trained model (Average
far-field surveillance videos, Autom. Constr. 85 (2018).
precision) AP50 = 60.08. To further improve the quality of the [14] J. Wu, N. Cai, W. Chen, H. Wang, G. Wang, Automatic
algorithm, it is planned to quantitatively and qualitatively detection of hardhats worn by construction personnel: a deep
increase the set of training data, and modify the architecture learning approach and benchmark dataset, Autom. Constr. 106
of the neural network. (2019).
[15] Z. Xie, H. Liu, Z. Li, Y. He, A convolutional neural network
REFERENCES based approach towards real-time hard hat detection, Proc.
[1] M.-W. Park, N. Elsafty, Z. Zhu, Hardhat-wearing detection for IEEE International Conference on Progress in Informatics and
enhancing on-site safety of construction workers, J. Constr. Computing (PIC), IEEE, pp. 430, (2018).
Eng. Manag. 141 (9) (2015). [16] N. D. Natha, A. H. Behzadanb, S. G. Paala, Deep learning for
[2] W. Liu, D. Anguelov, D. Erhan, C. Szegedy, S. Reed, C.-Y. site safety: Real-time detection of personal protective
Fu, A. C. Berg, SSD: Single Shot MultiBox Detector. In: Leibe equipment, Automation in Construction, Vol. 112, (2020).
B., Matas J., Sebe N., Welling M. (eds) Computer Vision – [17] Z. Cao, G. Hidalgo Martinez, T. Simon, S.-E. Wei, Y. A.
ECCV 2016. ECCV 2016. Lecture Notes in Computer Sheikh, OpenPose: Realtime Multi-Person 2D Pose
Science, vol 9905. Springer, Cham. (2016). Estimation using Part Affinity Fields, IEEE Trans Pattern
[3] J. Redmon, A. Farhadi, YOLOv3: An Incremental Transactions on Pattern Analysis and Machine Intelligence,
Improvement, (cite arxiv:1804.02767 Comment: Tech (2019).
Report), (2018).

13
 Incorporating Continuity between Professional
Subjects into the English Curriculum
Batbaatar Dashdemberel
Education department
Orkhon University
Ulaanbaatar, M ongolia
[email protected]

Abstract —The main goal of this work is to develop an

English curriculum in line with the content of specialized C. Objectives
design courses in order to improve the foreign language x Conduct a survey on professional English
teaching process, and based on this, to create an English language training in similar schools
textbook with a wide range of content, practical applied
exercises, and word chains. Based on some of the theories and x Develop a professional Eng lish design
methodologies of language teaching, students studying fashion curriculum
design using a text-based grammar-translation method to
develop a methodological framework for their application. x Develop professional English textbooks and
manuals based on the program
Keywords—Fashion, design, textbook, teaching, phrase
x Conduct experimental lessons according to the
curriculum
I. INT RODUCT ION
x Introduce innovative textbook versions into the
In the Mongolian higher education system, English is training, conduct experimental train ing and
taught as a general English course to students majoring in evaluate the results
other majors. However, a s mall nu mber of universities and
colleges are beginning to change their teaching methods to
include professional English. There is no consensus on the D. Research status of the topic
development and teaching of professional English Demid.B, (1998) “Imp roving fo reign language teaching
curriculu m yet some universities offer training based on their methodology based on the text” [16], Richards, J.C. (2001).
capabilities. Therefore, we need to exp lore opportunities to “Curriculu m Development in Language Teaching” [29],
learn and teach text-based English, and we need to explore Burmaa.Ts, (2011). “Methodological research to teach
opportunities to improve the effectiveness of professional English language reading skills to students” [13],
English training. Narantsetseg.M (2016) “Imp roving the learn ing process
using professionally-oriented Russian language textbooks”
A. Rationale [26], Mónica Rodríguez-Bonces, (2016) “A basis for the
design of a curriculu m incorporating music and drama in
There is a need to study the methodological issues of
children’s English language instruction” [25], Sirajul Munir,
improving the teaching process using professional-oriented
(2019) “Turn ing Classroom Pro ject into Textbook Design in
English textbooks for fashion design students that reflect
English for Specific Purposes Subject” [30].
modern needs. Therefore, the importance of this research
work is that the issue of designing English textbooks and the
curriculu m for undergraduate students majo ring in modeling E. Research methodology
is determined by the social significance and practical Test method is used to determine the level of knowledge
requirements. This study was used among students majo ring of students of the Fashion Design Institute with the content
in fashion design at the Fashion Design Institute (FDI) in of the new textbook
order to ensure that students in the professional field are Mathematical statistical methods are used to calculate
able to master the professional terminology and know the the results of experimental training
meaning and current news and trends. Experimental train ing on the new textbook version was
conducted using research analysis, summarization,
B. Purpose
comparison and observation methods
The main purpose of this work is to develop an English
language curriculu m in line with the content of specialized F. Scope of the study
design courses in order to improve the foreign language Conduct a survey on eight different universities and
teaching process. Based on this, to compile an English colleges that provide fashion design training
textbook with a wide range of design, appropriate exercises, The results of the experiment were p resented to the first,
and word chains, and to develop a methodological second- and third-year undergraduate students of the
framework for students majoring in fashion design based on Fashion Design Institute
some theoretical and pedagogical methods of teaching
foreign languages. G. Innovative aspects of research work
1) For the first time, studying and analyzing on the
curriculu m of professional design course and

14
 developed the new curriculu m o f Fashion 3. Urlakh Erdem Fashion Design Institute
Design course.
2) Co mpiled an innovative textbook with a wide 4. Mongolian State University of Arts and Culture
range of design texts, appropriate exercises, and
word chains based on the content. 5. City University

3) Introduced into teaching practice using

6. Ikh Zasag International University
professional English textbooks
4) 5 versions of new textbooks fo r English lessons 7. Fine Arts Design and Technology School
with codes ENLA-100[8], ENLA-101[9],
ENLA-102[10], ENLA -103[11], ENLA-
8. Rajiv Gandhi Polytechnic College of Arts and
104[12], were developed on a scientific basis,
and the results were evaluated Crafts

H. Practical significance of the research Provision of profe ssional

The School of Fine Arts, a branch of the University of te xtbooks and manuals
Culture and Arts, and Fashion Design Institute have been
involved in the development of English textbooks for 5% 24% Enough
students majoring in design.
Not enough
I. Research hypotheses 5 1% Uncertain
If the innovative English language textbooks for
vocational guidance are developed in a well-grounded and
scientifically sound manner, effectiveness and quality of the Fig. 2. Availability of professional textbooks.
training will be improved.
The purpose of chart 2 is to show the availability of
professional textbooks, newspapers and manuals resource in
II. EXPERIMENTS AND RESULT S English[4]. The graph shows a figure of 24% enough, 25%
not enough, and 51% uncertain, indicat ing that foreign
A. The graphic representation of the surveyed students language manuals, newspapers, and textbooks are
from all design universities is shown in the graph inadequate throughout the school, as well as to enable
students to conduct research and exp loration. Providing the
lib rary fund with professional books and textbooks on
Surveyed students from all design fashion design in English and Mongolian will facilitate their
universities research in this area, and it is necessary to create materials at
hand.
1
8 1 2
18% 10% 2 3 B. Lessons in a modern curriculum that meets the needs of
7 14%
4 students
9%
3 5 One of the most common methods used in foreign
6 9% 6 language teaching is text-based teaching, which is adapted
19%
5
4
to the learner's language level and selects sources that
7
14%
7%
present minor linguistic difficulties. Topic -based training is
8 unique in that it provides the advantage of using a variety of
usage terms.
Fig. 1. Average number of respondents.
1) Model Lesson-1 Types of Necklines.
Within the framework of this needs assessment survey, a a) Fill in each blank with a suitable WORD.
total of 154 students from 1st to 4th-year students majo ring
in fashion design and costume design at 8 universities and Blurry, bosom, cleavage, deeper, elegant, halter, hangover,
colleges in Ulaanbaatar were samp led using merits, plunging, striking
questionnaires[6]. The percentage of total school
participants was similar. This is considered to be an In wo men ’s fashion there are a lot of different types of
opportunity for all students majoring in fashion design, as neckline, each seeking to create a specific effect for the
well as for the universities and colleges where they are person wearing the dress, whether it's to increase the size of
trained, to facilitate further research. Surveyed students the ___________, to display varying amounts of cleavage or
from all design universities : simp ly to achieve a greater degree of comfort. In this guide
we explo re some of the more co mmonly used necklines and
1. Mongolia International University consider the ___________ of each. The Plunging Neckline;
This is a reasonably common neckline that is found on both
2. Mongolian-Korean Polytechnic College tops and dresses, as the name suggests it consists of a low -

15
cut area at the front which allows for varying amount of in groups. Between 10-15 new words are included to avoid
___________ to be displayed. When describing a duplicating data and taking less time[11].
___________ we prefer to add a little more detail and would
normally describe the shape as having a ‘V’, either a 'square' TABLE I. CROSSWORD-1 [11]
or a 'U' cut. The Scoop Neckline; Scoop necks are very
similar to a 'U' cut 'Plunging Neckline'; howev er, they № Crosswords’ questions-down
usually have a much ___________ cut and are not 2 more attractive than usual:
uncommonly found on things like T-Sh irts, with a recent 3 graceful and attractive in appearance or behavior:
trend making them very popular especially in men's fashion. 5 the narrow space between a woman’s breasts, that is seen when
A scoop neck tends to allow for more cleavage than a 'U' cut she wears a piece of clothing that does not cover the top of
plunging neckline would, but as you will appreciate where them:
we draw the distinction between the t wo can be a b it 7 a piece of women’s clothing that is held in position by a strap
___________. that goes behind the neck so that the upper back and shoulders
are not covered:
Scoop necklines are ideal for a slightly mo re flirtatious 8 a woman’s breasts:
outfit and are often seen in dresses aimed at the club wear
market. The Boat Neckline; A Boat Neck is a much wider TABLE II. CROSSWORD-2 [11]
neckline that is usually cut into a shallow 'U' shape which
№ Crosswords’ questions-across
often passes the collar bones and hangs on the shoulders.
1 the advantages something has compared to something else:
4 something that you cannot remember or understand clearly:
6 dropping suddenly or having a shape that drops a long way
down:
9 If a threat or doubt hangs over a place or a situation, it exists:
10 extending far down from the top or surface:

Fig. 3. T his picture is a part of the textbook [11].

It creates an ___________ look and is often seen in cocktail

and evening dresses. A very modest neckline that is ideal fo r
those situations where you want to show a bit of flesh but
where it is probably less appropriate to display large
amounts of cleavage. It is rarely seen in less formal attire.
The Off-the-Shoulder Neckline; Often referred to a Fig. 4. Find the words [11].
'strapless neckline' these necklines are similar to the 'Boat
Neckline' except they hang entirely off the shoulder. They c) Match the sentences, true or falseIncorporating
normally ___________ the arms below where the shoulder exercises to choose between true and false and matching
ends; however, when referred to as a 'strapless neckline' this sentences is intended to reinforce the basic mean ing and
would normally imp ly the neckline passes under the arms. informat ion in the text, as well as to encourage specialized
This is an ideal neckline fo r to accentuate the shoulders and vocabulary within the text[11].
neck of the wearing and often lends to a very elegant look.
[11]. TABLE III. MATCH THE SENTENCES [11]

№ Sentence-A № Sentence-B
b) Crosswords
1 T his is a reasonably 1 shoulders and neck of the
In addition to thinking skills, two types of word network common neckline that is wearing
exercises were introduced to develop visual and spatial found
skills, allowing students to work in pairs, indiv idually, and

16
 № Sentence-A № Sentence-B
2 Scoop necklines are ideal 2 where the shoulder ends
for a slightly
3 A very modest neckline 3 more flirtatious outfit
that is ideal for
4 T hey normally hangover 4 those situations
the arms below
5 T his is an ideal neckline to 5 on both tops and dresses
accentuate the

TABLE IV. T RUE /FALSE[11]

№ Put a tick () to show whether each sentence is T RUE T F

or FALSE.
1 A keyhole neckline can have two different meanings
2 Plunging neckline is rarely seen in less formal attire.
Fig. 5. Drawing with names [7].
3 A Boat Neck is a much wider neckline
4 Neckline has become increasingly uncommon.
5 Scoop necks are very similar to a ‘U’ cut

d) Write the name of pictures from the text

Includes exercises for working with pictures that fit
professional vocabulary to stimu late activit ies such as
composing sentences based on pictures, creating dialogue,
and comparing new words in text with pictures. When a
teacher walks through a task and identifies a common
mistake made by a student, the teacher can work closely
with the student and encourage them if the student tries to
find the mistake but is unable to do so. Students 'inspiration
fro m the teacher and the teacher's progress in the students' Fig. 6. Drawing with names [7].
skills, as well as the interesting completion of the exercises,
According to the professional English curriculu m, word
are the main mot ivations for the teacher to prepare well fo r
the lesson [11]. ordering exercises are included to teach students to spell
new professional terminology correctly.
TABLE V. W RITE THE NAME[11]
Boosm, Mietrs, Cvgaelae, Pgluning, Deeepr, Blrury,
№ Write the name of pictures from the text Elgeant, Hvaengor, Hltear, Srtiikng
1 2 ________________________________________________
________________________________________________

In addition to drawing according to the professional

English program, the students also made small-scale works,
__________________ which became a great experience in their sewing, design and
__________________
embro idery techniques[7]. Satisfaction research clearly
3 4 shows that students and teachers , became part of a training
program that was very inspiring in terms of creativ ity and
practice.

__________________ __________________

e) Unscramble the words

Fig. 7. Making small-scale of works [7].

17
 There are a test sample of a min i-shape of a skirt design,
which is widely used by wo men, to create a unique image o f 3 68.96
the garment. As a result of the course, innovative works are
emerging. 2 54.34

1 63.7
2) Memorize the words while drawing and making
small-scale of work Control class 70.2

In an experiment to draw a picture of what is expressed 0 20 40 60 80

in the vocabulary within a given topic to memorize
professional vocabulary, 20% o f the students answered that
Fig. 10. T he average grade of three years of study evaluation-2.
it was high residual, 13% easy to memorize, 3% d ifficu lt,
57% interesting, and 7% moderate. 90% of the respondents In the 2011-2013 academic year, English language
said that giving them a task to memo rize pro fessional training was provided based on the translation of short texts
vocabulary in accordance with their skills allo ws them to also without any professional manuals, only info rmation
learn and motivate. was collected fro m the online environment and the words
the professional terminology were exp lained to the students
using a projector[6]. The graphic shows that the results are
Memorize the words while drawing very low, only 1.24% higher than average. The average
grade of C indicates the need to develop and introduce
professional textbooks.
7 0 High residual
20
Easy to memorize 5) Average grade point for the 2012-2014 academic
Difficult year [6].
13
Interesting
57 3 Moderate.
4 57.43
Uncertain

3 83.2
Fig. 8. Memorize the words [7].
2 63.69
3) Integrate English lessons with other professional
curriculum 1 67.5

0 20 40 60 80 100
Like it very much
Fig .11. T he average grade of three years of study evaluation-1.
Like it
5%
0% Experiments conducted in the 2012-2014 school year
11% Like it moderately showed that the average grade of the control class was
48% 14.54% higher than the highest grade of the normal class [6].
Little bit This suggests that it is mo re effective to use professional
36%
terminology in English lessons, and to use new words in
Dislike
sentences and dialogues in the form of ro le-p lays. It was
also 11.39% more effective than the average for the 2011-
2013 school year.
Fig. 9. Integrate English into professional curriculum.
1) Average English language assessment for the first
This graph shows that 48% like it very much, 36% like term of the 2019-2020 academic year
it, 11% like it moderately, and 5% a little bit. Th is leads to
the conclusion that in the future, it is necessary to integrate
English into professional courses at universities.

4) Average grade point for the 2011-2013 academic

year [6].

18
 98 emotional support for the publication of Fashion design
96 English textbooks. Above all I would like to express my
deep gratitude to Professor D.Nyamjav, Ph.D., Sc.D who
94
mentored and supported me in challenging times.
92
90 96.5
88 95 94.6 92.7 93.8 93.8 93.6 REFERENCES
86 90.4 90.3 [1] Ayar, Z. & Yangın-Ekşi, Y. (2019). “Continuing professional
84 87.1 87.1 development options of foreign language school directors”,
International Online Journal of Education and Teaching (IOJET),
82 6(3). 707-721.
1,1 1,2 1,3 1,4 2,1 2,2 2,3 2,4 3,1 3,2 3,5 [2] Azhar, S. M. & Shahid, A. (2014). “ Sketching effective faculty
professional development framework”, Asian Journal of Business
Management, 6(2), 118–123.
Fig. 12. T he average grade of the last semester of study evaluation [3] Baird, , Alexander, (1976), “The study and teaching of literature”,
ELT J, 30, 4, 221–6.Google Scholar
In the first half of the year, the average score was 92.2% [4] Batbaatar.D (2012) "Opportunity to motivate students to learn English
when the Eng lish language curriculu m was developed in using professional learning materials" unpublished report
conjunction with the professional curricula of the university, [5] Batbaatar.D (2012) "Orientation to Teaching English to Professional
Students" unpublished report
and textbooks and manuals were developed and tested. This
[6] Batbaatar.D (2014) “ Opportunities to introduce fashion design
graph shows a dramatic change in the quality of education. English to English training”, Research paper №14, Orkhon
University, 2014
[7] Batbaatar.D (2016) “ Methodological issues of closely linking the
content of professional English courses with other curricula” (on the
III. CONCLUSION example of professional design), Research paper №16, Orkhon
University 2016
Based on our research on “Incorporating Continuity [8] Batbaatar.D, Fashion design English-1, UB 2016
between Professional Subjects into the English Curriculu m” , [9] Batbaatar.D, Fashion design English-2, UB 2016
we developed a detailed curriculu m in conjunction with [10] Batbaatar.D, Fashion design English-3, UB 2016
other professional curricula in collaboration with
[11] Batbaatar.D, Fashion design English-4, UB 2016
professional teachers of design theory and methodology.
[12] Batbaatar.D, Fashion design English-5, UB 2016
Experiments and research on professional English teaching
[13] Batmunkh.T s and Erdenemaam.S, “Theory and methodology of
methods have been successful. This is one of the most foreign language teaching”, UB, 2014
important issues for professional English teachers. [14] Batmunkh.T s, “Theory and methodology of teaching foreign
languages”, UB, 2008
We have developed an English textbook for [15] Bazarragchaa.B, “Comprehension of language”, UB, 2004
undergraduate students majoring in fashion design. The [16] Burmaa.T s and Bayanjargal P, “Methodology of Teaching English”,
possibility of teaching English based on textbooks was Ulaanbaatar, 2012
studied, as a result, professional English language training [17] Burmaa.T s, “Methodological research to teach English language
provided them with the professional vocabulary of design reading skills to students”, Doctoral dissertation, UB, 2011
and laid the foundation for research. [18] Chuluundorj.B, “Introduction to the study of Mongolian language in
parallel with non-native language”, UB, 1999
[19] “ Collaborative Learning ISS Proceedings”, 2013, Page 9
It is necessary to further improve the English language [20] Demid.B, “ Improving foreign language teaching methodology based
teaching methodology for professional fashion design and on the text”, Doctoral dissertation, UB, 1998.
introduce it in university curricula. Educational institutions [21] Galsan.S, “ Linguistics and other aspects”, UB, 2008
and communit ies that provide train ing not only in the field [22] Gurbazar.R, “ Fundamentals of translation theory and practice”, UB,
of design, but also in other innovative fields, should 1996
regularly develop textbooks and manuals that contain the [23] Hajidsuren.N, “ Principles of foreign language teaching”, UB, 1998
content and concepts of the innovative field. [24] Ichinhorloo.Sh, ”Training methodology reform”, UB, 2009
[25] Khishigdelger.B, “Foreign language teaching methods, theory and
Creat ive assignments by students in accordance with the application”, UB, 2015
professional English curricu lu m had a positive effect on the [26] Luvsandorj.T s, “Curriculum development and educational standards”,
learning outcomes and created the conditions for innovative UB, 2004
work. Professional English training is more focused on the [27] Mira.N and Burmaa.Ts, “Problems of English language education and
solutions”, UB, 2009
pragmatis m than theory, and we believe that this
professional English training is effective and extremely [28] Mira.N, “Theoretical and methodological issues of teaching English
to secondary school students”, Doctoral dissertation in Education,
useful. UB, 2003
[29] Mónica Rodríguez-Bonces, “ A basis for the design of a curriculum
incorporating music and drama in children’s English language
ACKNO WLEDGMENT instruction”, Issues in Teachers’ Professional Development, 19(2),
203-223.
I exp ress my gratitude to B. Mungunkhulug, Ph.D., [30] Narantsetseg.M (2016) “Improving the learning process using
Associate Professor, who encouraged me to do research professionally-oriented Russian language textbooks”, Ph.D. thesis
work, G. Tungalag Ph. D, Director of the Mongolian [31] Nyamjav.D, “Theory and methodology of foreign language teaching”,
Fashion Design Institute, who provided financial and UB, 2009

19
[32] Oyun.B, “Exploring the possibility of using game methods in Russian [35] Sukhomlinsky VA, “ One hundred tips for teachers”, UB, 1985
language teaching in secondary schools”, Doctoral dissertation in [36] T sedev.Kh, “Ways to increase professional terminology in foreign
Education, UB, 2006 language lessons”, Page 180
[33] Richards, J.C. (2001). "Curriculum Development in Language [37] T uya.Sh, “Modern methods and technologies of foreign language
T eaching", Cambridge: Cambridge University Press. teaching”, UB, 2010
[34] Sirajul Munir, “Turning Classroom Project into Textbook Design in
English for Specific Purposes Subject”, Dinamika ilmu, Volume
19(1), 2019

20
 Education in Pandemic Crisis Implementation via
ICT Online Technologies
Cyryl Koshyk
Director of Technology, AI & Global Expansion ( ICT & Innovation Department)
The Kyoto College of Graduate Studies for Informatics
Kyoto, Japan
[email protected]

Abstract—Due to recent unpredicted COVID-19 pandemic, In 20th century education has changed rapidly with the
the worldwide global outbreak halted the world still. Education development of learning technologies.
sectors were deeply disrupted and tremendously affected. The
situation, forced nearly all schools and universities to rapidly Technologies can now deliver easy to access, proactive
adopt new ICT technologies, shifting into online technologies, and comprehensive learning and teaching environments.
adopting them partially or fully. This new modality, provided
distance learning and teaching, called Education 2.0. II. EDUCATION 2.0
Education 2.0 is online distributed learning for students,
regardless of their physical location which may be affected by As of 2020, students have higher expectations, on
isolation restrictions and country lockdowns. Time critical integrating ICT into classrooms.
assimilation of ICT into education sector became Many researches has proven that usage of ICT in the
indispensable, in order to adhere side effects of pandemic process of learning and teaching, can greatly improve
outbreak of COVID-19.
students achievements, as well as technology based
In Japan, at The Kyoto College of Graduate Studies for education can be useful in – increasing creativity and
Informatics, actions had to be taken rapidly regarding thinking skills.
COVID-19, creating never tested before large scaled successful New generations are born and have grown up with these
online ecosystem, that provided uninterrupted online based technologies. They are defined as “digital natives”. As a
education platform consisting of an all in one learning result, the younger the students, the tendency for them to
education system. This included complete distance learning have a much higher interest and demand for strong ICT
and teaching technologies as: scoring, attendance reporting,
integration.
multi language support, direct interaction as group or one-on-
one, advanced students monitoring. Current online systems provide easy, intuitive learning
factors, giving confidence to professors, and engaging
Keywords—ICT, Pandemic, Education, Artificial students more into achieving greater accomplishments.
Intelligence, Distant Learning, eLearning
Professors express high confidence and competency in
I. INTRODUCTION using ICT to create deeper integrated forms in their
classrooms. When used well, content can become richer, and
Distribution method for education has changed forever, allow for better visual explanations. ICT has to power to
the aftermath of COVID-19 (Coronavirus) will never be offer state of the art interaction, and is now available even on
forgotten, preparing us today for similar, pandemic outbreaks portable mobile devices such as – phones or tablets.
in the near future, where ICT will be crucial and break major
ground for all education sectors.
III. ADOPTED TECHNOLOGIES
ICT (Integration of Information, Communication, and Portable mobile device hardware and processor
Technology) has long assisted school and universities - staff, advancements, allow for the possibility of highly interactive
teachers and professors, to replace traditional teaching learning without the need of personal computers. At KCG
methods with technology-based teaching and learning tools (The Kyoto College of Graduate Studies for Informatics) we
and facilities. While this process has been slow, at the have fully adopted various online platforms, systems and
beginning of 2020, the escalating and exponential infections services which are fully supported on mobile devices, if no
of COVID-19 (Coronavirus) worldwide, forced all schools computers are present.
and universities to resume partial or full operations via ICT
and specifically usage of online platforms, teachers Current following technologies and platforms were used
professors and school stuff had to quickly learn such in online distant teaching and virtual classrooms:
systems, no matter their backgrounds or knowledge. By
x Artificial Intelligence auto translation services
utilizing and adoption of ICT, education has been resumed.
provided by Google - ML AI
In Japan, ICT is considered as one of the main elements,
transforming the country into future development, with a x DeepL (www.deepl.com) for human like AI
strong focus on online eLearning systems and platforms. translated text content for online courses.
Technology-based importance in learning and teaching it is x Zoom (www.zoom.us) for live interaction and
becoming more present in schools. attendance monitoring

21
 x YouTube Live (www.youtube.com) backboned by OBS rendered output in form of “Virtual Web Camera”
Google cloud servers, for live 4k streaming could be feed directly into any streaming software or online
platform, as a single video feed at any given resolution (4k,
x Slack (www.slack.com) and Discord HD).
(www.discord.com) for mass interactive multi chat
with students To improve and provide clear audio quality, teachers
used mobile phones as microphones.
x UNASENSE (www.unasense.com) Cutting edge,
worlds first Artificial Intelligence online platform WO Mic (www.wolicheng.com) app and software was
specifically designed for education used as virtual microphones. From in depth testing, we
concluded that, mobile phone microphones provided superior
x dotCampus LMS (www.dotcampus.com) for quality with very low audio noise compared to even higher
publishing, posting, archiving content end webcams built-in microphones.
x Custom written PHP powered modules for interactive
quiz, tests and monitoring, served by Linux operating V. CONTENT QUALITY CONTROL
systems in the cloud, provided and hosted by Google To deliver the highest quality content possible where
Cloud. sound, verbal and visual communication were key factors of
x CDN (Content Delivery Networks) provided by distant teaching, equipment such as camera and microphone
CloudFlare (www.cloudflare.com), for fastest were taken into consideration. We tested multiple hardware
possible delivery to countries outside of Japan, and devices and concluded mobile devices, specifically
ensuring steady and fast hosted content. phones delivered the best quality, overall portability, and
were already distributed to all faculty staff.
By mixing all these technologies together, a new learning
system was provided for complete uninterrupted teaching Technically benchmarked [1], mobile phones are
and learning solution. equipped with much higher CMOS imaging sensors, then
normal computer or laptop cameras. Where 4k and even 8k
resolutions can be achieved, rendering high quality image
IV. CONTENT PREPARATION quality, making it a main choice in our eLearning course
For content preparation, screen capturing software was teachings.
used for easy technical adoption for all professors
Performing well in low light conditions, mobile
(irrespective of age groups or ability with screen
microphone sensitivity delivers high quality sound. Some
capturing/streaming knowledge. Software such as Screen-O-
teachers also used Bluetooth devices for more portability.
Matic (www.screencast-o-matic.com), and FastStone Screen
Recorder (www.faststone.org) were employed. All performed well, and are highly suggested for any type
of video/audio content recordings and online live streaming.
Professors having access to PC based stationary
computers or laptops equipped with Windows 10,
additionally offered already built-it XBOX Game Bar VI. INTERNET REQUIREMENTS
feature, allow for high quality offline video content Many students without access to computer labs relayed
preparation, without the need of additional software. solely on their own mobile devices (phones, tablets). All the
Professors were trained in groups and one on one, via online school’s eLearning content was accessible even on lower
tutorials both in Japanese and English (multi language). bandwidth networks such as 2g and 3g.
All school members, staff and professors were allocated 4k high resolution video streaming feed was
small space in their locations, used green screen material automatically downgraded to lower resolutions, depending
backdrop mounted on wall or tripods, in order to display on viewing device, 4k for regular broadband internet (Cable),
background elements on screen. “Green screen” is a known HD 1920x1080 for 4g, LTE, and 1280x720 for 3g networks
technique, normally used in TV stations for broadcasting as and lower.
well as in feature films productions of visual effects,
allowing green screen material to be replaced and instantly All live video content while streaming was also recorded
turning any wall into fully “transparent backdrop”. This and storage locally on teachers drive. This was, later
allowed for static pictures or animated videos to be displayed uploaded and archived for offline viewing into dotCampus
directly behind the speaker. online eLearning LMS platform.
Powerful multi language open source software OBS
VII. CLASSROOM INTERACTION
(Open Broadcaster Software www.obsproject.com) was used
to create virtual set, basic or complex containing multi- Virtual classrooms greatly differ from traditional ones
layered visual elements such as - pictures, charts, videos, (real presence in physical classrooms).
giving more control of content placement, compared to
However, interesting patterns could be observed in
standalone direct feed web camera.
virtual classrooms. Isolated “single-out” students seem to
OBS software platform enabled teachers to include live focus better, as they felt more responsible and more in
websites with interaction and browsing capabilities, specific control, while in their own private environments, students
full screen or cropped software screen grabs, interactive were less vulnerable, stressed and more open to discussion.
widgets, live playback YouTube reference videos, Better performing students were interacting and active
PowerPoint Charts and more. more in online chat rooms, delivering higher quality home-
work and assignments.

22
 All students class hours were tracked by the online YouTube Live, which enabled live auto translations into
platform dotCampus LMS; email and phone push different languages. In The Kyoto College of Graduate
notifications were sent to inform them of next upcoming Studies for Informatics, international students were able to
lessons. watch video streaming in their native language. Additionally,
AI based DeepL services was used for preparing
E-mail communication didn't satisfy our expectations. downloadable PowerPoint presentations in multi languages.
Students equipped with personal computers tend to have
response that were usually delayed. We shifted into mobile All live streaming content after online airing had
devices, which were key in active and fast responsive embedded subtitles.
communication.
By using Google ML speech to text [2], audio spoken
Teachers and students used mobile apps for mass multi language could be embedded into the video stream as well.
chat instant communications such as Slack and Discord at all
times. Students were assigned to specific class groups, few Artificial intelligence has dominated speech to text and
contributing students were also assigned as administrators to text to speech engines for years now, it is a great extension to
filter chat content, and monitor all chat activities. any eLearning platforms.

Additionally, the Discord app allowed for voice and X. ATTENDANCE MONITORING CHALLENGE
webcam communication.
During the pandemic, where the online world was the
Online hours lesson were done via Zoom software app only safe way of communication with students, eLearning
for live feed streaming. class attendant monitoring has been a challenge, whereby
Students using laptops or personal computers could share students were not physically present in classrooms.
their screens directly with teachers, if help or support was The Kyoto College of Graduate Studies for Informatics
needed. Teacher could access and control the student designed its own proprietary system for attendance
keyboard and mouse, helping in specific tasks. monitoring. Students logged to the learning system, using
Many students used both mobile devices for direct their student unique ID number.
communication such as video, voice and chat and personal Additionally, in some classes, "Face ID" AI system were
computer online or offline as work machines. applied, by using custom built artificial intelligence based -
In mandatory courses and classrooms exceeding 250 face recognition system. Some students didn't have access to
students, to provide highest quality streaming, we used computer webcams, so mobile devices (phones, tablets) were
YouTube Live, with outputted low latency in video used instead.
streaming delivery. At most a 5 second delay was noted. As mobile devices are multi-sensory devices, equipped
From our testing, we conclude that YouTube Live is a with cameras and microphones, "Face ID" AI system directly
solid platform for mass streaming. If one-way video scanned students faces from their phone, and recorded online
streaming is sufficient for specific lessons, performance was attendance.
superior and stable, and able to handle hundreds to thousands In conclusion, mobile devices are most versatile form for
of students at once. Additionally YouTube Live offers up to attendance monitoring no matter location or internet
8 hours nonstop recording directly to YouTube, backed by bandwidth speed.
Google servers, downloadable, with built in on the fly auto
translation into over 30 languages.
XI. ASSIGMENTS DELIVERY AND ONLINE GRADING
After completing their assignments, students directly
VIII. ARTIFICIAL INTELLIGENCE FOR EDUCATION
uploaded their assignments in the form of digital files into
The Kyoto College of Graduate Studies for Informatics is dotCampus LMS online platform at specified deadlines.
the world first university to fully utilize and adopt it's own, Teachers could then directly download the files as a single
100% online based artificial intelligence platform ZIP file containing all student assignments and review them.
UNASENSE (www.unasense.com). The system which was
originally designed and started by Prof. Cyryl Koshyk. This Assignment uploading is supported by all devices in all
powerful and flexible system is able to create standalone forms personal computers and mobile devices (phones,
mobile AI systems and programs made by students. It was tablets). In case of revisions, or additional comments,
first tested for medical AI programs and content, detecting teachers were able to log directly to student’ computers via
skin cancer on mobile devices, at Etugen University ZOOM or TeamViewer, encrypted bi-directional platforms,
(www.etugen.edu.mn) in Mongolia. and communicate directly with students in one on one
sessions.
UNASENSE AI has been deeply integrated into school
operations and routines, including assessment of student
XII. CONLUSSION AND FEEDBACK
knowledge and skill level. The platform will be available to
universities and schools around world, after all features has We performed an in depth student questionnaire online
been added and final testing phase has been conducted and evaluation. Our outcome reports showed that 76% students
complete. found eLearning distant learning compared to traditional
learning to be very good or good. Less then 5% of students
found it technically difficult however they learned quickly
IX. MULTI LANGUAGE SUPPORT
how to use such systems.
Due international students’ presence, live content was
prepared in multi-language release and streamed by Google's

23
 Overall, we found that online courses and classes can “Virtual office hours” were provided to all students.
provide a high engagement rate. Professors, and staff specified hours and days where they can
be contacted online, for direct one on one communication for
Students primary preferred choice of accessing eLearning students to seek advisory, support and ask questions.
content was by using mobile devices.
Covid-19 (Coronavirus) has challenged education.
Strong focus should be given, to visual format of Students may prefer online over physical schools, where
delivery, and content optimized for slower internet (mobile artificial intelligence has been adopted and utilized. In 2021
3g,4g,LTE) for such devices. Most of our video content was we all will experience Education 2.0. More possibilities will
delivered in vertical screen format, which allowed for exist by using technology, resulting in new rules,
simultaneous live chat communication while video streaming requirements, and new policies. The future edu has begun
was performed. not as a choice, but as a path of transformation.
Vertical and square format videos are well suited for
eLearning delivery into mobile devices. REFERENCES
Teachers using mobile devices for their high-quality [1] Veli-Tapani Peltoketo, “Mobile phone camera benchmarking:
cameras and microphones render superior quality over most Combination of camera speed and image quality”, DOI:
10.1117/12.2034348, January 2014.
personal computer webcams; it also makes content more
[2] Tim Ashwell, J.R. Elam, “How accurately can the Google Web
portable. Speech API recognize and transcribe Japanese L2 English learners’
The learning curve is not complex both for teachers or oral production”, DOI: 10.29140/jaltcall.v13n1.212, 1892, April
2017.
students. We found that within a week, online eLearning
systems can be learned, allowing students to navigate, post
content, and enable overall interaction.

24
 Risks in the Educational Sector: Theory and
Practice
Enkhtuvshin Jambaldorj
Sector for higher and professional education research
Mongolian Institute for Educational Research
Ulaanbaatar, Mongolia
[email protected]

Abstract - This study explains the relationship indicate that the Mongolian education sector is in a risky
between academic achievement and individual, family, situation. Identifying the root causes of such consequences
and social life, identifies the causes of risks in the (risks) from the system is not only a challenge for the
education sector, considers risks in relation to education sector but also for Mongolian society, but to
philosophical categories and systematic laws, develops reduce risks in the education sector, protect against potential
risk identification and classification models, and risks and according to V.V. Razuvaev reduce “the likelihood
formulates educational risks using this model. The issues of unintended consequences” [2] and improving the output
related to the sector policy will be evaluated in relation and performance of the education sector has become a
to the research results and suggestions on what to do current need.
next will be considered.
A. Goals
Keywords: cause-effect, risk, system, risk Contribute to the development of risk mitigation
classification, risks in education, risk management in measures in Mongolia's education sector.
education.
B. Objectives
I. INTRODUCTION
1) Compare the theoretical formulas proposed by
The academic performance of Mongolian secondary researchers on risk;
school students is not in the level of satisfactory. For 2) Formulate risks and highlight specifics in the
example: education sector;
ƒ According to the international TIMMS and PIRSL 3) Identify the root causes of educational risk;
scores, primary education is 37.8-41.5 percent and 4) Compare theoretical formulas with practical
basic education is 22.9-28.5 percent, which is 7 examples;
percent lower than the international average; [10] 5) Propose options for implementation at the national
ƒ In the general entrance exam held in 2019 only 28-37 level;
percent performance in subjects with the largest C. Literature review
number of examinees (Mongolian language,
Mathematics and English); [11] The concept of risk dates back hundreds of years and has
ƒ According to the quality assessment of secondary the root back to antiquity. ”In the sixteenth and sixteenth
education (Ministry of Education, Culture and centuries, the basic concepts of opportunity and probability
Science. 2019), 63 out of 100 examiners have shown were developed, as well as gambling, in the seventeenth and
dissatisfied results; [11] eighteenth centuries, the theory of probability and statistics,
in the nineteenth and twentieth centuries, progression and
These outcomes, which are the first gateway to the regression analysis. From the beginning of the twentieth
education system, continue to have a negative impact not century, when the theory of the game emerged, the concept
only on the quality of secondary education, but also on of a new modern risk was formed, researchers say. [13] In
youth employment, quality of life, economic and social the 1930s, A. Marshall and A. Piguo created a neoclassical
development, and competitiveness. It is still attracting flow of risk in the economic field, intensifying theoretical
attention that in “2019 the Mongolia's competitiveness ... research.
ranks last in terms of indicators directly related to the
education sector, such as skilled engineers, pollution issues, In our country, banking and insurance companies have
higher education, and knowledge exchange” [12] “There is been implementing risk management and risk research since
currently no publicly accepted understanding of risk in the mid-1990s, and government implementing agencies
education. Often, the risk is seen as a manifestation of have been implementing risk management since 2006. The
uncertainty, damage to the operation of an educational Education Inspection Agency (Education Inspection Agency
institution, or an undesirable situation or outcome”. [1] All is part of the government) uses checklists to identify risks
the data above, the researcher's conclusion and position associated with the material environment of secondary

25
schools and kindergartens. Apart from this, no risk studies ƒ Uncertain consequences of actions or events related
have been conducted in the education sector. to human values;
ƒ A combination of opportunity and loss; [2]
Information on risk theory and methodology is often
sourced from external sources, and government agencies Social risk is defined as “deviation from the norms of
include consulting services from the Institute for Risk basic social indicators for individuals, groups, and society as
Management in the United Kingdom (IRM), the Association a whole, and the basic indicators include unemployment,
of Insurance and Risk Managers (AIRMIC), and the aging, death, disability, orphanhood, and divorce”. The
National Forum for Public Risk Management (ALARM) survey results are:
Organizations. In addition, the ISO GUIDE 73: 2002 Risk ƒ Risk is subject to (1) cause-and-effect patterns and
Management - Vocabulary standard issued by the systemic laws, (2) probabilistic in nature;
International Organization for Standardization has been used ƒ Risk is a time-dependent continuous phenomenon
as a reference document. The standard (This standard is that takes on a positive, negative or zero value at a
accepted in many countries, especially in all EU countries, given moment;
including Russia, and in the Commonwealth of Independent ƒ Social sector risks are considered in relation to basic
States, such as Canada and Australia) states that risk is a social indicators;
combination of both the probability of an event (occurrence) ƒ Educational risks in Mongolia are largely
and its consequences, both positive and negative (Risk unexplored.
Management Standard © AIRMIC, ALARM, IRM: 2002,
D. Methodology
translation copyright GDNT: 2007). The social sectors
consider risks depending on the nature of their functions and Research should follow as Colin Robson puts it, "to
activities under the law. For example: build a detials and in-depth knowledge of one or a few
interrelated examples." [5] and exemplified research
ƒ In many areas, such as security and the environment,
strategies and system approaches. As the procedure;
“negative” “consequences” are defined as risks
(International standards state that this is only the case 1) Collect examples and analyse evidence related to
in the security sector); output, outcome, and impact in education system;
ƒ In the legal field, “the likelihood of an increase in 2) To study the theoretical formulas proposed by
potential adverse effects”; researchers on risk;
ƒ In the field of civil law, “damage caused by natural 3) Explain risks in terms of cause-and-effect, systematic
and man-made disasters and dangers, regardless of patterns, and develop a classification model;
human will [3]; 4) Identify educational risks based on the model;
ƒ In the public service sector, “risk is the occurrence of 5) Explain and evaluate the education system and the
a situation in which a service cannot be performed or model of personal and social development in terms of risk
is impossible to be fulfilled in accordance with a theory;
planned or obligation”, In the field of state control, 6) Formulate ways and solutions to reduce risks in the
“risk means that the failure to adequately fulfil the education sector.
responsibilities of state control imposed by law will II. RESULTS
lead to social, economic damage and other adverse
From the findings of the survey:
effects”, In the field of environment, it is stated that
“there is a possibility of adverse effects on animals, ƒ Causes are the effects of systemic influences;
plants, human health and the environment” ƒ The process of cause and effect is continuous;
(Introduction to the United States Agency for ƒ The state of the system at the moment is both a cause
International Development (USAID) Economic and a consequence;
Reform Project); ƒ It is logical to assume that the process of causation is
In the economic sphere, risk is defined as “risk that subject to the law of transition from quantity to
includes ... events that may or may not occur. When this quality.
happens, there are three economic consequences: negative The above-mentioned philosophical concepts will help
(defeat, loss, failure), zero, positive (win, gain, profit)”. [4] to explain risk in education.
Professor V. V. Razuvaev, a member of the Russian
National Academy of Economics, discusses many aspects in The Internationally Recognized Survey (TIMSS), at the
his book. For instance the risk is: national level, shows that students' academic achievement is
very poor, indicating a risky situation in the education
ƒ Percentage of potential losses and losses and sector. The knowledge, skills, and attitudes of learners are
damage; the underlying causes of the quality of life, happiness, and
ƒ Probability and extent of unintended consequences; development of a country, as “risk” itself is the cause of
ƒ A combination of probabilities and consequences; other consequences. According to researchers, “lack of
ƒ A combination of three scenarios, the probability of a knowledge and skills is the cause of poor human resources,
scenario and its consequences; poor human resources are the cause of poor wages, and poor
ƒ A two-way combination of "event-consequence" and wages are the cause of poor living conditions, on the other
its associated "uncertainty"; hand, general culture and moral weakness (risk), poor
physical development. The risk of all aspects of human

26
development, these two risks create the risk of not being TABLE 2. RISK CLASSIFICATION MODEL
able to take their rightful place in society” [6]. Effects (output)
Reason Process
(size,
Students' knowledge, skills and attitudes are the (input) (transformation)
amplitude)
(probabilities)
primary cause of civic development on the one hand, and a
means of adapting and communicating with the external Рh
Ro+s Тt
environment (family, school, community) on the other. In Еh
this regard, the theoretical conclusion is that “everything
that exists contains a clear signal about its nature, origin and Тs Рm
environment ...” [7] Every living thing adapts to Rs
endangered environments, whether they exist or not, Еm
develop or perish. Therefore, the evolution (development) of
Тm
any individual (In Buddhism, all living beings are Рl
considered) is determined by its ability to adapt to a risky Ro
Еl
environment. Most of our students have not learned such
Тf
skills. What is the reason for this?
Key factors influencing the outcome of the education
system are teachers, programs, learning technologies, and
budgetary and financial inputs. It is the responsibility of the This includes:
state to compile, develop, allocate, and improve efficiency. Ro- objective reason
The state makes decisions (policies, laws) and fulfills its Rs- subjective reason
responsibilities. Therefore, education policies and decisions Ro+s - combined (objective and subjective) reason
are the starting point and key variable that affects academic Tt - time of the process
achievement. Using these inputs in the education system, the Ts - space of the process
learner undergoes a qualitative transformation, creating the Tm - ability to manage the process
following “consequences” for society. Tf - manifestation of the process
1) Learner's knowledge, skills, and attitudes toward Eh - large amplitude
output. Em- medium amplitude
2) The quality of life of the citizen (family) on the El - small amplitude
outcome. Ph - high probability of occurrence
3) The country's level of development and Pm- medium probability of occurrence
competitiveness on impact. Pl - little probability of occurrence
To summarize:
In this model, educational risk is defined as Ro+s → [Tt +
TABLE 1. EDUCATION SYSTEM MODEL Ts + Tm + Tf] → [Eh] → [Th], expressed in words:
Transformation
1) for a combination of reasons;
Input Output Outcome Impact 2) medium and long-term exposure;
process
Reason Process Effects 3) latent;
The Changes in Changes in Change 4) the magnitude of the consequences is large
Education sector (insurmountable);
student individual the family in society
5) manageable;
When the operation of the system brings positive results 6) Is a high probability social risk.
at every stage, a knowledgeable and capable citizen is As mentioned above, “public policy decision”, which is
developed, the well-being of the citizen is ensured, and thus the starting point of the system, is a subjective cause but a
the well-being of the society is ensured. Otherwise, the decisive one. Let's look at two related facts.
guarantee of the well-being of the "citizen" will be lost and
the country will fall into decline. This correlation suggests 1) In the last 30 years:
that the primary cause of positive and negative ƒ Education law 12 times
consequences in any society is the state of education and the ƒ 6 times the structure and type of school
education system (risk situation). Therefore, researching and ƒ Primary and secondary education standards 4 times
diagnosing sector risks and preventing potential threats is a ƒ training program 5 times
policy issue. The education sector is a complex system that ƒ the curriculum 11 times have been changed [8]
covers a wide range of social life under the influence of 2) In the field of education:
many external and internal factors. Therefore, a risk ƒ Guidelines for government change in the education
identification and classification model has been proposed sector -1
and an attempt has been made to identify industry risks ƒ Master plan -1
accordingly. ƒ National program - 11
ƒ Sub-program - 3
ƒ Sector development step-by-step program - 2 was
approved and implemented [14]

27
 A sufficient number of policy documents have been IV. CONCLUSION
adopted and a sufficient number of changes have been
1) To establish a professional, independent “institution”
made. But, there is no independent body or human resources
of risk management independent of the executive, which
to analyze these conflicting decisions. “Risk awareness can
exercises national oversight of education, reports to
be performed by external consultants, but a well-
parliament;
communicated, regular, and well-organized internal
2) Establish an independent policy research unit within
approach is likely to be more effective. It is essential that the
the structure of educational research and methodological
risk management process be owned within the
institutions and increase research capacity.
organization”. (Risk Management Standard © AIRMIC,
ALARM, IRM: 2002, translation copyright GDNT: 2007)
Unfortunately, this international principle has not been REFERENCES
implemented. [1] R.A.Rasulya, “Problems of risk research in education”,
III. DISCUSSION Scientific article. vol 3, Kazan RA Federal University, Russia,
ƒ Risks in the Mongolian education system have 2016, pp 2.
become one of the real causes of risks for Mongolian [2] V. V. Razuvaev, “Theoretical definition of the concept of risk”
families and Mongolian society; // Modern scientific research and innovation, URL:
http://web.snauka.ru/issues/2016/09/72007.
ƒ Loss of continuity of government policy on
education due to political instability; foreign [3] S. Narangerel, “Legal dictionary”, Ulaanbaatar: 1 rd ed, 2007,
pp 626.
education was simply imported; There is reason to
believe that scientific research and undiagnosed risk- [4] B. Tserendorzh, “Risk management” Ulaanbaatar. T&U
making policies and decisions have created a risk printing LLC, 2008, Page 11.
situation in the education sector. There is research [5] Colin Robson, “Real word research”, (1993) Copyright © 2006
evidence to support this. "One-third of the population "Monsudar", Open Society Forum,Ulaanbaatar, pp 42.
is nomadic herders, and one-third is poor and [6] P.E. Shcheglov, N.Sh.Nikitina, “The quality of higher
extremely poor, but much of Mongolia's education education. Risks in the training of specialists”, University
Management. 2003, vol 24, pp 46-59.
reform is based on the experience of a high-income,
sedentary country. Mongolian policymakers are [7] D.Byambasuren,” The wheel of the universe” Ulaanbaatar:
Urlakh erdem LLC, 3rd ed, 2009, pp 30.
actively involved in the policy lending process” [9];
ƒ According to the agent's theory (the theory that, the [8] J.Enkhtuvshin, “On the reform of the education management
system” 2017, Institute of Education. Research paper. vol 22.
body that oversees the implementation of
pp 43.
parliamentary decisions is independent of the
[9] Gita Steiner-Khamsi, Ines Stolpe. “Educational Import: Local
government) there is no independent entity to
Encounters with Global Forces in Mongolia”, Ulaanbaatar.
monitor the implementation of decisions on Admon, 2003, pp 2.
education issues in the interests of all citizens;
[10] Minister of Education, Culture and Science, “Concept of
ƒ Improving the capacity of Mongolia's policy research quality reform of primary and secondary education”, 2013
institutions on education and establishing and https://www.meds.gov.mn/.
developing a risk management and control system is [11] Educational Evaluation Center under the Ministry of
a pressing issue; Education and Science http://stat.eec.mn/.
ƒ Risks in the education sector need to be studied in [12] Centre for Economic Policy and Competitiveness Research.
https://ecrc.mn/.
depth at the institutional level.
[13] Historical study of the development of risk management
concepts. Electronic conference. http://econf.rae.ru/ .
[14] Government of Mongolia, 2010, National Education Program
of Mongolia, 2010, https://www.legalinfo.mn/ .

28
 Sample Model for Selecting State Examination
Test Item
Examination test items in Social studies for the 12th grade students of Secondary School

Gan-Erdene Jigjid Ganbat Tsend

Education Evaluation Center School of Information and Communication Technology,
of the Ministry of Education and Science Mongolian University of Science and Technology
Ulaanbaatar, Mongolia Ulaanbaaatar, Mongolia
[email protected] [email protected]

Abstract—The Education Evaluation Center develops assess- A normalization in Mongolian language is not discussed here.
ment to assess the knowledge and skills of secondary school Normalization is consisting of stemming and lemmatization
students and takes state exams every year, and currently it has [1]. You can find it in our previous article on word normal-
a database of 23 subjects and 32000 tests. Select test item from
the test database with expert’s help and takes that year’s state ization published in Mongolian [4], [5].
exam. To select test items that aren’t similar or same as last Normalize source test item by using dictionary will be as
year’s is a very time-consuming process. In addition, a test item follows. Following tasks are reported in Mongolian due to
is not expected to be repeated in two consecutive examinations. usage of mongolian vocabulary:
This paper aims to present the experiment results we obtained
in developing machine models to reduce human participation, D1 = иргэн сурах өмнө хооллох хувцаслах орон-
to determine task similarity and to select different tasks from байр байх ёс
a database and to implement them in item database system of D2 = төр иргэн оронбайр өмчлөх
12th grade social studies students of the Secondary Schools. D3 = иргэн ажилгүйдэл төр багасах үүрэг
Keywords — weight of a term, test item similarity, exper- D4 = нийлүүлэлт багасах бараа үнэ өсөх
iment, vector, evaluation D5 = илүүдэл нийлүүлэлт үнэ багасах
I. D ICTIONARY DEVELOPMENT English translations:
D1 = citizen study before feed clothe apartment be
5 types of tasks in Social studies include. Following tasks
D2 = government citizen apartment owns
are reported in Mongolian due to usage of Mongolian vocab-
D3 = citizen unemployment government decrease re-
ulary [2] . A task is means that a sample question of test unit
sponsibility
except answers. .
D4 = supply decrease commodity price increase
D1 = иргэд суралцахаас өмнө хооллож, хувцас- D5 = excess supply price decrease
лаж, орон байртай байх ёстой
Let’s consider a task as document.
D2 = төрөөс иргэдэд орон сууц өмчлүүлсэн
D3 = иргэдийн ажилгүйдлыг төрөөс багасгах үү- II. D ESCRIBING TASKS AS VECTORS
рэгтэй
The term frequency T F of term t in document d is defined
D4 = нийлүүлэлт багасвал барааны үнэ өсдөг
as the number of times that t occurs in d, (T F > 0) [1].
D5 = илүүдэл нийлүүлэлт үнэ багасгадаг
Construct incidence matrix by term frequency in task (see
English translation: the Table II). DF - is the document frequency, the number
D1 = citizens must fed, clothed and housed before they of documents that t occurs in. IDF - is log weight of term. It
study is found by the following formula [1]:
D2 = the Government has privatized the apartments to
citizens N
IDF = log10
D3 = the Government has to bear the responsibility to DF
reduce unemployment of citizens N is the number of documents in the collection (in our
D4 = when supply decreases, commodity prices will case, collection is a test database or test). After that the log
increase frequency weight of term is found by the formula:
D5 = excess supply will cause price to fall
T F W = 1 + log2 T F (T F W ≥ 1)
A normalization can be used to extract a word stem from a
test unit to create a corresponding word dictionary. Table I See the Table III.

28

29
 Table I
D ICTIONARY

Term ID Dictionary term /Mongolian/ Source terms /Mongolian/ Dictionary term /English/ Source terms /English/
T1 Ажилгүйдэл Ажилгүйдлыг unemployment unemployment
T2 Багасах Багасвал, багасдаг decrease reduce, decreases, fall
T3 Байх байх be be
T4 Бараа Барааны Commodity commodity
T5 Ёстой Ёстой be must
T6 Илүүдэл илүүдэл Excess excess
T7 Иргэн Иргэд, иргэдэд,иргэдийн Citizen Citizens, to citizens, of citizens
T8 Нийлүүлэлт Нийлүүлэлт Supply supply
T9 оронбайр Оронбайр, оронбайртай, оронсууц House housed, apartments
T 10 Өмнө өмнө Before before
T 11 Өмчлөх Өмчлүүлсэн own Privatized
T 12 Өсөх Өсдөг increase Increase
T 13 Сурах Суралцахаас study study
T 14 Төр Төрөөс Government government
T 15 Үнэ Үнэ Price prices
T 16 Үүрэгтэй Үүрэгтэй Responsibility Bear the responsibility
T 17 Хооллох Хооллож feed Fed
T 18 Хувцаслах хувцаслаж clothe Clothed

Table II
I NCIDENCE M ATRIX
By finding the weight of a term as IDF is multiplied by
T F W (Table IV) each task is expressed as a real-valued vector
TF d1 d2 d3 d4 d5 DF IDF [3]. Each task is now presented as vector. In a vector space
t1 1 1 0.70
t2 1 1 1 3 0.22 Table IV
t3 1 1 0.70 W EIGHT OF A TASK ’ S T ERM
t4 1 1 0.70
t5 1 1 0.70 TW d1 d2 d3 d4 d5
t6 1 1 0.70 t1 0.00 0.00 0.70 0.00 0.00
t7 1 1 1 3 0.22 t2 0.00 0.00 0.22 0.22 0.22
t8 1 1 2 0.40 t3 0.70 0.00 0.00 0.00 0.00
t9 1 1 2 0.40 t4 0.00 0.00 0.00 0.70 0.00
t10 1 1 0.70 t5 0.70 0.00 0.00 0.00 0.00
t11 1 1 0.70 t6 0.00 0.00 0.00 0.00 0.70
t12 1 1 0.70 t7 0.22 0.22 0.22 0.00 0.00
t13 1 1 0.70 t8 0.00 0.00 0.00 0.40 0.40
t14 1 1 2 0.40 t9 0.40 0.40 0.00 0.00 0.00
t15 1 1 2 0.40 t10 0.70 0.00 0.00 0.00 0.00
t16 1 1 0.70 t11 0.00 0.70 0.00 0.00 0.00
t17 1 1 0.70 t12 0.00 0.00 0.00 0.70 0.00
t18 1 1 0.70 t13 0.70 0.00 0.00 0.00 0.00
a In our case, d is a task. t14 0.00 0.40 0.40 0.00 0.00
t15 0.00 0.00 0.00 0.40 0.40
t16 0.00 0.00 0.70 0.00 0.00
Table III t17 0.70 0.00 0.00 0.00 0.00
log F REQUENCY W EIGHT OF T ERM t18 0.70 0.00 0.00 0.00 0.00
length 1.772 0.924 1.111 1.159 0.924
TFW d1 d2 d3 d4 d5
t1 1
t2 1 1 1 with 18 axes of t1 · · · t18, documents d1 to d5 in Table IV,
t3 1
t4 1 will be the coordinate vector
t5 1 D1 = (0, 0, 0.7, 0, 0.7, 0, 0.22, 0, 0.4, 0.7, 0, 0, 0.7, 0, 0, 0,
t6 1
t7 1 1 1
0.7, 0.7)
t8 1 1 D2 = (0, 0, 0, 0, 0, 0, 0.22, 0, 0.4, 0, 0.7, 0, 0, 0.4, 0, 0, 0, 0)
t9 1 1 D3 = (0.7, 0.22, 0, 0, 0, 0, 0.22, 0, 0, 0, 0, 0, 0, 0.4, 0, 0.7, 0,
t10 1
t11 1
0)
t12 1 D4 = (0, 0.22, 0, 0.7, 0, 0, 0, 0.4, 0, 0, 0, 0.7, 0, 0, 0.4, 0, 0, 0)
t13 1 D5 = (0, 0.22, 0, 0, 0, 0.7, 0, 0.4, 0, 0, 0, 0, 0, 0, 0.4, 0, 0, 0)
t14 1 1
t15 1 1 In the last row of a Table, there are length of vectors. The task
t16 1 length is found by follow formula
t17 1 
t18 1
|Di | = d2i i = 1, 5

29

30
 Table V
III. C OMPUTING COSINE SIMILARITY OF TASK C OSINE S IMILARITY OF D OCUMENTS
Since each task is represented by a vector in a vector space
∼ d1 d2 d3 d4 d5 Average
model, cosine similarity is measured by the cosine of the d1 1 0.127 0.025 0 0 0.0379
angle between two vectors. The smaller the angle between d2 0.127 1 0.202 0 0 0.0822
two vectors, the more similar are the vectors, and the less d3 0.025 0.202 1 0.038 0.048 0.0783
d4 0 0 0.038 1 0.342 0.095
similar otherwise. The formula to find the angle between two d5 0 0 0.048 0.342 1 0.0974
vectors is shown below [3]. Average of d1 to d5 0.0782

n
Ai · Bi
cosine similarity = cos(θ) =  i=1  least similar compared to d1 and d2. The content of the tasks
 n  n
  also shows the difference. The cosine averages of the given 5
 A2 ·  B2 i i tasks can be ranked in ascending order as follows. (Table VI)
i=1 i=1
Table VI
Where, A, B- different tasks. Ai , Bi - its coordinates R ANK O F D OCUMENTS
Since each document is similar to itself, its cosine similarity
is 1. If the cosine value is closer to 0, the two tasks’ similarity Rank Average Document similarity
1 0.0379 d1 The least similar
will be less. The similarity of documents has been found using 2 0.0822 d2
the cosine formula given below. (See the Table V) 3 0.0783 d3
The three tasks d1 , d2 and d3 with the words t1 and t2 in the 4 0.095 d4
5 0.0974 d5 The most similar
case describe the state exam in vector space, a If the average is close to 1, it is similar to the others.
Conversely, closer to zero is different from others.

IV. S OME RESULT OF EXPERIMENT AND CONCLUSION

When we used this model to develop cosine similarity
of tasks program to calculate cosine similarity from social
studies’ 1881 tasks, cosine similarity of test database of tasks
was calculated as 0.741 [6]. Also, state exam tasks should be
less similar, and the cosine similarity should be to closer to
zero. The cosine similarity of test database after removing 800
similar tasks was 0.13. This sample model will be the most
Figure 1. A state exam consisting of many similar tasks suitable to evaluate the performances of similarity measures of
selected N state examination task documents and new tasks
added to the test database, to make a list of similar tasks, to
add the least similar tasks to a test database and to make task
clustering by chapter or etc.
R EFERENCES
[1] Christopher D Manning, Prabhakar Raghavan, Hinrich Schutze, "Deter-
mining the vocabulary of terms, Scoring, term weighting and the vector
space model," in Introduction to Information Retrieval, UK: Cambridge
University press, 2008, pp. 69, 146, 155, 158.
[2] Education Evaluation Center, "Sample test of documents," [Online].
Available: http://itembank.eec.mn/home.
[3] Ganbat TS, Sarangerel D, “Fundamental of Information Retieval in
Mongolian”, MUST edition, Ulaanbaatar, 2019, pp. 116, 121.
[4] Gan-Erdene J, Ganbat Ts, "Experiment results of Mongol stemming
algoritms in Mongolian," Technical report of the project to improve the
Figure 2. A state exam consisting of less similar tasks quality of tasks by Machine Learning Method, Education Evaluation
Center, Ulaanbaatar, January 2020.
[5] Gan-Erdene J, Ganbat Ts, "Term normalization in Mongolian," Technical
The average value of cosine similarity between other tasks report of the project to improve the quality of tasks by Machine Learning
is given in the last column of the Table. The cosine similarity Method, Education Evaluation Center, Ulaanbaatar, January, 2020.
[6] Gan-Erdene J, Ser-Od B, Ganbat Ts "Evaluating Heap’s law constans
of given documents can be evaluated by the overall average based on a document database in Mongolian," in Proceedings of MMT
of separately averaged values. Cosine similarity of the given 2020, Ulaanbaatar.
5 test item-based database is 0.0782, cosine similarity for d4
and d5 is 0.342 which means that they are the most similar
against each other. It is clearly seen that d4 and d5 are the

30

31
 Perception of Lane Curvature Detection for
Situational Awareness
Julkar Nine Shanmugapriyan Manoharan Mohammad Didar Hossain Sagar
Department of Computer Engineering M.Sc.Embedded Systems M.Sc. Automotive Software Engineering
Technische Universität Chemnitz Technische Universität Chemnitz Technische Universität Chemnitz
Chemnitz, Germany Chemnitz, Germany Chemnitz, Germany
julkar.nine shanmugapriyan.manoharan mohammad-didar-hossain.sagar
@informatik.tu-chemnitz.de @s2017.tu-chemnitz.de @mb.tu-chemnitz.de

Shadi Saleh Wolfram Hardt

Department of Software Engineering Department of Computer Engineering
Technische Universität Chemnitz Technische Universität Chemnitz
Chemnitz, Germany Chemnitz, Germany
shadi.saleh wolfram.hardt
@informatik.tu-chemnitz.de @informatik.tu-chemnitz.de

Abstract—Lane curvature detection from the highway is Traffic light and traffic sign detection, pedestrian detection,
one of the fundamental requirements for both autonomous lane tracking, lane curvature detection, etc. are some
driving and advanced driver assistance (ADAS) based ap- example applications in this area that can be combined
plications. Curvy lane is a crucial factor that plays a
significant role in the increasing number of road fatalities to build a smart application like driving assistance and
worldwide. Detection of the curvature radius of the lane self-driving car. Lane departure is a prime factor for most
and providing a warning message for upcoming curvy road of the highway fatalities that causes loss of lives and
conditions have the potential to improve driving safety. This money every year. Lane departure is the reason for more
paper presents an efficient algorithm for lane curvature than 50% of the total car crashes happens every year in
detection combining multiple image processing techniques.
The proposed algorithm has been implemented in Python and the USA [3]. Not following the curve on the highway
evaluated on a Raspberry Pi 3 B+ environment. It can detect lanes properly is the basic cause for departing from the
and calculate lane curvature as well as provide warning highway. Lane curvature detection has the potential to
messages for curvy road conditions. provide a solution in this aspect. This is what motivates
Index Terms—Raspberry Pi, Binary Thresholding, Gradi- the automotive researchers to research on this topic. An
ent, Bird’s-eye View.
efficient lane curvature detection can bless us with the
solution for road safety. In this paper, an implementation of
I. I NTRODUCTION
an image processing algorithms to detect and recognize the
Car production and purchase have increased to a high Lane Curvature for situation awareness using a monocular
extent in the last few years to satisfy the increasing camera was developed. It focuses on Binary Thresholding,
transportation demand of the world demography. The Bird-Eye view Transformation and Sliding window to
automotive industry is expecting to sell 74 million cars detect the circles of traffic light signs. The implementation
in 2020 worldwide [1]. With this increasing number of was tested on a Raspberry Pi 3B+.
automobiles on the roads and highways, a huge deteriora-
tion of traffic management and safety are being observed. II. LITERATURE REVIEW
Number car crashes and road other fatalities are rising A. Situation Awareness
to a high degree. In 2017 over 21000 fatal car crashes In Mica R. Endsley’s study [4], the three levels of
happened in the USA. Intelligent Transport System (ITS) situation awareness were introduced: perception, compre-
[2] is the modern technology that works for safety and hension, and projection. Where the information from the
traffic perfection by providing automation in vehicular environment perceived by the perception level, compre-
communication. Many countries are planning and devel- hension level understands this information, and projection
oping various ITS applications to improve transportation. projects a future action based on the taken action. In this
Advanced driver assistance system (ADAS) is an example situation, Lane curvature recognition will be a useful case
of ITS which supports drivers in driving, parking, etc. study in the perception level of Situation Awareness.
ADAS also helps drivers by proving warning messages
about rough road conditions, curvy lanes, traffic signs, and B. Lane Curvature Recognition
speed limits, etc. Other than that, the self-driving car is Different methods have been offered in both machine
also being considered as a solution to traffic problems and vision and computer vision. Systems that have been de-
safety issues. But some dependencies must be satisfied veloped so far can perform various tasks such as lane
with the successful development of these applications. mark detection, lane departure warning, lane curvature
32
detection, lane-keeping assistance, etc. Researchers from III. C ONCEPT
Toyota Central Lab proposed a lane detection method This paper presents the concept and implementation
for highway lane departure warning using an extended process of Lane Curvature Detection using Image Pro-
Hough Transform [5]. Extended Hough transform helped cessing. The monocular camera feeds the input to the
to improve the robustness in lane detection. Hardware cost image processing algorithm to detect the lane curvature.
is reduced because they used a wide-angle camera and The proposed method of lane curvature detection consists
an embedded CPU available in the market. Lee J.W. [6], of several steps that include image processing techniques,
presents a machine vision lane departure detection system as shown in the flow diagram (Fig. 1). The proposed
using a unique edge distribution system. Noise effect algorithm is a pipeline with several steps such as Binary
and localization error are minimized with this function. Thresholding, Bird’s-eye View Transformation, Sliding
Edge related information and lane related information is Window to Locate Lanes, and Lane Curvature Detection.
collected and combined by edge detection function (EDF)
to identify the traveling direction. Park J.W. et al. [7],
proposed an algorithm for lane curve detection using
a Lane Curve Function (LCF). The proposed algorithm
was tested on a video of a Korean superhighway and
showed excellent results even in noisy frames. It is worth
mentioning that, a robust lane curvature detection method
has been proposed comparing the lane line’s slope with
the desired slope in [8]. Their target is to support driver
assistance and autonomous driving. The proposed method
works well in adverse conditions such as noise, lane mark
discontinuity, and disturbances. It performs shadow and
illumination correction and differentiates between different
curves. The algorithm can be used in autonomous vehicles
for lane detection as well as on normal vehicles for a speed
control system. Jeon Y.W. [9], offers an image processing
method for preventing lane deviation. By processing and
compressing the captured image from the road, slope, and
curvature of the lane can be determined in this method.
Driving direction and can also be detected. Our internship
also focuses on lane curvature detection using image pro-
cessing without any learning algorithm. From a conceptual
point of view, their goal is partially related to ours. Both
Goldbeck J. et al. [10] and McCall J.C. [11], focus on
lane detection and tracking. The system in [10] uses a
highly dynamic CMOS video camera for real-time lane
detection with high accuracy. It can adapt to any weather
condition and road type without external adjustment. In Fig. 1: Flow Chart of Lane Curvature Algorithm
[11] on the other hand, works for lane mark detection
using a steerable later. It shows robustness to lighting The proposed algorithm is implemented on the Rasp-
changes and shadow effects. This method performs with berry Pi 3B+. The operating system used in the Raspberry
accuracy in detecting circular reflector road marks as well Pi is Raspbian Stretch. The libraries such as OpenCV,
as painted lane marks. Zou Q. et al. [12] proposes a Numpy are installed on Raspberry Pi with Python depen-
hybrid deep architecture by combining the convolutional dencies for performing image processing tasks. OpenCV’s
neural network (CNN) and the recurrent neural network getPerspectiveTransform function is used to transform the
(RNN) by investigating frames from a continuous driving image into the bird’s eye view. NumPy’s polyfit function
scene. Apart from these, there are other researches related fits a second-order polynomial and returns the coefficients.
to our work. Jung C.R. et al. [13], is a linear parabolic
method for lane following. Jung C.R. et al. [14], is a IV. IMPLEMENTATION
similar but improved approach by the same researchers. This work focused on the recognition of lane curvature
which includes curve detection as well. A lane detection from sequences of frames captured from a monocular
method to support the lane-keeping system is offered camera mounted on the vehicle. Firstly, the threshold-
in [15] by estimating the lane configuration and vehicle ing techniques such as color thresholding and gradient
attitude. Tsogas M. et al. [16], supports the lane-keeping thresholding are applied to the input frame, to get the
systems using the digital map. In this study, we focus binary image. The lane lines are visible in the binary
on a combination of Binary Thresholding, Bird-Eye view image. In color thresholding, the S-Channel from the HLS
Transformation and Sliding window for recognizing Lane color channel makes the lane line more visible. Using the
Curvature. gradient thresholding steep edges can be detected which
is most probably indicates lanes. The Sobel filter and
thresholding applied to the pixel value that falls inside the
33
range of minimum and maximum range. Then, the output The equation can be simplified as below:
of the color and gradient thresholding are combined to
[1 + (2ay + b)2 ]3/2
obtain the binary image. Rcurvature = (3)
|2a|
Furthermore, a warning message will be generated
based on the lane curvature. If the curve is inclined to
the direction left, the warning message should be ‘Curve
to the left’. If the curve is inclined to the right message
should be ‘Curve to the right’. The value of the polynomial
coefficients analyzed such that a relational pattern is ob-
served to decide whether the curvature is inclined towards
left or right.
V. R ESULTS
The proposed algorithm achieves the goals of this
Fig. 2: Lane Curvature Recogntion Algorithm
internship. The algorithm is implemented in Python3 and
Secondly, using OpenCV’s getPerspectiveTransform, tested on a Raspberry Pi environment. It can detect the
the binary image is transformed into a bird’s eye view. lane curvature from a highway video. It shows a warning
getPerspectiveTransform function accepts four sources and message while there is a curvy lane moving to the right
four destination points. In general, lane lines are parallel or left. Fig. 3 shows the result for both ‘curve to the left’
lines and they meet at a vanishing point that forms like and ‘curve to the right’ respectively.
a triangular shape. The tuning is applied to get the four
source points for the Region of Interest (ROI) which avoids
the region other than lane. Similarly, the destination points
also selected to form a rectangle shape. The getPerspec-
tiveTransform takes the source and destination points as
input and provides the bird’s eye view image of the region
of interest polygon.
Thirdly, plotting a histogram will give two peaks on two
different positions for left and right lane lines. These two
peak points are the starting points for a sliding window
search to locate both left and right lanes. Sliding windows
are rectangles drawn over the pixel points of the left
and right lanes. The size and number of the window are
selected manually according to the shape of the image. The Fig. 3: Lane Curvature Recognition: Left Curvature → Radius
two windows are set at the bottom of the image on the two to left and right lane curvature is 217m and 201m; curve to
x-coordinate positions indicating to the left and right lanes. the left (top), Right Curvature → Radius to left and right lane
That means, two x-coordinates of the histogram peaks are curvature is 392m and 245m; curve to the right (bottom)
two centers for two base windows. If the number of pixels
inside a previous window (left or right) is greater than Blue text on the result frames shows the radius of
a fixed threshold the base position for the next window curvature for the left lane and the red text shows the radius
(x-axis) is updated on their mean position. If it does not of curvature for the right lane. The green text refers to
exceed the threshold windows are not recentered. This way the warning message. Since the first original frame in the
window on both the left and right sides follows the left figure has a curvy lane moving to the left, our system is
and right lane lines respectively. Connecting the centers showing the warning message ‘Curve to the left’. On the
of these windows on both sides provides two curves that other, the second frame has a curve in the right direction,
refer to the left and right lane. so the message says, ‘Curve to the right’. This algorithm
Finally, the lane coordinates are stored in the list of can detect lane curvature accurately even in shadow, noise,
python programming. The lane lines from the bird’s eye and less lighting conditions. Video frames were converted
view image are near vertical. It is more likely to have the to the binary image before other processing techniques are
same x-coordinate for multiple y-coordinate. In this case, applied. Applying HLS S-Channel helps with this situation
it is a better idea to fit a 2nd order polynomial for f(y): by making the binary image less affected by the lightness
and shadow. This algorithm works best on a 480p video.
x = f (y) = ay 2 + by + c (1) The proposed algorithm can process a 480p video with 25
NumPy polyfit function fitts a 2nd polynomial and FPS in an average on a windows platform. On the other
returns the polynomial coefficients a, b, and c. The radius hand, raspberry can process the video with a speed of
of curvature at any point x of the function x = f(y) is 15-20FPS. The algorithm shows similarly accurate results
calculated from the equation below: for 720p video but the processing is lower. The algorithm
dx 2 3/2
calculates the radius of curvature accurately. In some rare
[1 + dy ] cases, it shows a false positive decision for ‘Curve to the
Rcurvature = d2 x
(2)
dy 2
left’. Thresholding on the value of polynomial coefficients
34
(a, b, and c) is tried to x the problem. It lowers the error [5] A. Takahashi, Y. Ninomiya, M. Ohta, M. Nishida, M. Takayama:
rate but cannot solve it entirely. Rear View Lane Detection by Wide Angle Camera. In: Intelligent
Vehicle Symposium, 2002. IEEE. vol. 1, pp. 148-153. IEEE (2002).
Our algorithm fails to detect lane curvature in some [6] J.W. Lee: A Machine Vision System for Lane-Departure Detection.
special cases such as multi-lane conditions, dividing one Computer vision and image understanding 86(1), 52-78 (2002).
lane into two or more, sudden narrow or wide lanes [7] J.W. Park, J.W. Lee, K.Y. Jhang: A Lane-Curve Detection based
on an LCF. Pattern Recognition Letters 24(14), 2301-2313 (2003).
appearance, etc. In such cases, sometimes neighboring [8] A. Dubey, K. Bhurchandi: Robust and Real time Detection of Curvy
lanes are detected which is not the desired outcome. To Lanes (curves) with Desired Slopes for Driving Assistance and
handle this problem our system can take a pause when the Autonomous Vehicles. arXiv preprint arXiv:1501.03124 (2015).
[9] Y.W. Jeon: Image Processing Method for Preventing Lane Devia-
road condition is critical to detect lane and resume it after tion (Feb 1 2005), uS Patent 6,850,629.
a few frames. [10] J. Goldbeck, B. Huertgen: Lane Detection and Tracking by Video
Sensors. In: Proceedings 199 IEEE/IEEJ/JSAI International Confer-
ence on Intelligent Transportation Systems (Cat. No. 99TH8383).
pp. 74-79. IEEE (1999).
[11] J.C. McCall, M.M. Trivedi: An Integrated, Robust Approach to
Lane Marking Detection and Lane Tracking. In: IEEE Intelligent
Vehicles Symposium, 2004. pp. 533-537. IEEE (2004).
[12] Q. Zou, H.Jiang, Q. Dai, Y. Yue, L. Chen, Q. Wang: Robust Lane
Detection from Continuous Driving Scenes using Deep Neural
Fig. 4: Lane Curvature Detection in Low Light Condition → Networks. IEEE Transactions on Vehicular Technology (2019).
Radius to left and right lane curvature is 3702m and 3796m; [13] C.R. Jung, C.R. Kelber: A Robust Linear-Parabolic Model for
curve to the right Lane Following. In: Proceedings. 17th Brazilian Symposium on
Computer Graphics and Image Processing. pp. 72-79. IEEE (2004).
[14] C.R. Jung, C.R. Kelber: An Improved Linear-Parabolic Model
for Lane Following and Curve Detection. In: XVIII Brazilian
Symposium on Computer Graphics and Image Processing (SIB-
GRAPI’05). pp. 131-138. IEEE (2005).
[15] Tsuji, M., Shirato, R., Furusho, H., Akutagawa, K.: Estimation of
Road Configuration and Vehicle Attitude by Lane Detection for a
Lane-Keeping System. SAE transactions pp. 420-426 (2001).
[16] M. Tsogas, A. Polychronopoulos, A. Amditis: Using Digital Maps
to Enhance Lane Keeping Support Systems. In: 2007 IEEE Intel-
ligent Vehicles Symposium. pp. 148-153. IEEE (2007).
Fig. 5: Lane Curvature Detection in Critical Road Condition:
Radius to left and right lane curvature is 330m and 493m; curve
to the right (left), Radius to left and right lane curvature is 818m
and 943m (right)

VI. CONCLUSIONS & F UTURE W ORK

Curvy lane and lane departure are the reasons for many
car fatalities on the highway. Accurately detecting lane
curvature and providing warning messages for curvy road
conditions can support many Intelligent Transport System
(ITS) applications. The proposed algorithm in this report
has been evaluated on a Raspberry Pi environment and it
has shown success in lane curvature detection as well as
a warning message. Analyzing the 2nd order polynomial
coefficients from the curve fitting to decide the warning
message is a novel idea to reduce a lot of computation
complexity. This algorithm works fine in shadow and low
light area. Combining this algorithm with other computer
vision algorithms that concerns autonomous driving and
driver assistance can develop a large-scale ITS application.
The future plan is to detect the current lane from a multi-
lane highway. Other than that, detecting multiple lanes
from the highway and showing warning messages for lane
departure is also part of the future work.

R EFERENCES
[1] Statista-Research-Department: Global-car-sales-1990-2019 (June
2020), https://www.statista.com/statistics/200002/ international-car-
sales-since-1990/
[2] I. Wagner: U.s. fatal passenger car crashes 2017 (April
2020), https://www.statista.com/statistics/191530/ fatal-passenger-
car-crashes-in-the-us/
[3] G. Lu: A Lane Detection, Tracking and Recognition System for
Smart Vehicles. Ph.D. thesis, University of Ottawa (2015).
[4] M. R. Endsley “Toward a Theory of Situation Awareness in
Dynamic Systems”, Human Factors Journal 37(1), 32-64, 1995.

35
 MRI Rendering with EEG Sensors Projection
Klementev Dmitrii
Computer Engineering Department
Novosibirsk State Technical University
Novosibirsk, Russia
[email protected]

II. MRI PROCESS

Abstract — The article describes MRI (Magnetic Resonance MRI imaging process is based on hydrogen nuclei in a
Imaging or Magnetic Resonance Investigation) visualization by
human body and its spin, a powerful magnetic field, and a
Direct Volume Ray-Casting rendering method and the
principle of the MRI process. The latter is important for weak oscillation magnetic field [1]. In the powerful
understanding all circumstances of MRI visualization. For the magnetic field, which magnetic isolines are almost straight
high performance of the calculations, GPU (Graphics in the workspace, nuclei are aligning parallel to the field
Processing Unit) was used. DirectX was used to link CPU and with very small relaxation time. In other words, nuclei spins
GPU. EEG (Electroencephalogram) is a method of brain are aligned parallel to the isolines of the powerful magnetic
electrical activity investigation. The method of projecting EEG field. The weak magnetic field begins to oscillate
data to the tissue was described. This method can be used to perpendicular to the powerful stationary magnetic field with
project any data. resonance frequency. With the oscillations, nuclei begin to
precess around the powerful magnetic field. Such resonance
Keywords— Magnetic resonance imaging, direct volume
rendering, real-time rendering (computer graphics), medical frequency is called precess frequency and depends on a
visualization. magnetic field magnitude. At 1 tesla this frequency is 42
MHz. After nuclei have been brought away from
I. INTRODUCTION equilibrium, the weak magnetic field is turned off.
Eventually, nuclei spin will align with the powerful
MRI is one of the non-invasive methods that allow magnetic field, but it takes some time. Meanwhile, radio
us to investigate the human body. Preventing unnecessary waves at this frequency are emitted from the body. Those
surgery, this aspect is very important for medical radio waves can be measured.
diagnostics and researches. In general, there are two similar The question here is how to define from what
approaches for the non-invasive investigations: places the radio waves have been detected on sensors. Paul
1. Using X-rays (Computed Tomography scan) Lauterbur has introduced a method, which allows
2. Using magnetic fields (Magnetic Resonance determining nuclei positions [1]. His idea was to make the
Investigation) powerful magnetic field inhomogeneous and to use the
effects of it. For that, he used three different linear magnetic
The CT (Computed Tomography scan) provides fields, perpendicular to each other, along with X, Y, Z for
density information of tissues, MRI provides other example. There are three steps:
information, based on the radio-wave response, getting with 1. Precess frequency depends on the magnetic field
magnetic fields. The information can be about hydrogen magnitude. If you create a linear magnetic field,
density, neuron activity, or brown motion of particles. The only nuclei on one isoline will have the same
MRI has a very strong advantage: it doesn’t use X-rays, that precess frequency. This step is choosing a slice and
could damage the body. This gives MRI a good reputation. this magnetic field is called “slice encoding”
The problem of MRI rendering is the fact that 2. The next applied magnetic field is called “phase
radio-frequency response is not optical properties and encoding”. It is turned on for a short time. During
different tissues can be indistinguishable in this field. For that time nuclei rotate in the plane perpendicular to
mapping MRI scalar data, representing radio-frequency “phase encoding magnetic field”. As the frequency
response and saved as 3D texture, there are transfer of that rotation is dependent on magnetic field
functions. It makes transfer function generation an important magnitude and the magnetic field is linear, each
part of MRI visualization. nucleus will have a personal alignment depending
The grey matter of a human brain is a major on the nuclear position along the magnetic field. As
component of the central nervous system and it includes the magnetic field is turned on for a short time,
numeral neuronal cell bodies, while white matter includes nuclei will not change precess frequency along
few cell bodies, but a lot of long-range myelinated axons. “slice encoding magnetic field” and radio waves
The gray matter is found on the surface of the brain and is frequency will be the same.
relatively thin. By this fact, sensors on the brain can 3. It gives the X slice. With the second step, we can
measure neuron activity. This is performed by EEG. The determine nuclei along Z. The last step is to
neuron activity can be investigated for getting information determine nuclei along Y and by that take row
about processes in the brain. Usually, EEG is represented by values. The last magnetic field is changing nuclei
numbers or diagrams. The mapping of the EEG data relative frequency along Y and we can determine row
to the brain could increase understanding of processes in the taking radio waves with only one frequency.
brain because visualization is one of the most important
communications between a human and nature.

36
 In the end, we have a row of nuclei with the same
frequency and each nucleus has its phase.
An MRI image is a three-dimensional digital field,
each of its cells represents an averaged nuclei response in
the corresponding physical space as a scalar value [1]. Such
cells are called “voxels” like “pixels” for a two-dimensional
space. Values in those voxels will be called as “scalar data”
in this paper.
III. MRI VISUALIZATION
Fig. 2. (a) is cube rendering with back polygons only and (b) is cube
Direct Volume Ray-Casting Rendering rendering with front polygons only

Direct Volume Rendering is a ray-casting method High performance of rendering is in demand. The
for calculating the color of a screen pixel by casting a ray imaging can be calculated on CPU and GPU, but GPU is
from the pixel through a three-dimensional texture, such as much more effective due to architecture for parallel
an MRI image [2]. In a simple implementation, each voxel calculating and a possibility to calculate pixels color in
has color and opacity parameters. By these parameters, the SIMD style. For that purpose, DirectX11 can be used. It
color of the screen pixel can be calculated (1,2) [2]. includes pixel and computes shaders, each of these can be
used for calculations.
(1) To improve sensitivity, it is possible to shade
opacity tissues. In general, normals can be represented as a
(2) normalized minus gradient of scalar data. So it can be
calculated with the formula below (1):
Where C is the color and A is the opacity. The
opacity value is used for color blending between the volume
and other objects. The opacity number is between 0 and 1
and the strength of the opacity is higher at 1.
But using these equations requires to be careful
with choosing the color because color values with low (4)
opacity can make render white like in Fig. 1 because of the
excess of color value. Where grad is the gradient of scalar data, gradi is
the i-component of the gradient, h is the distance between
the voxels along the i axis, n is the normal.
This calculation is expensive due to six sample
calls. The good practice is to save calculated normals into an
additional texture before the main render loop. It will
discard additional sample operators and leave only one. In
shading calculations, it can be important to set threshold
value, because of high noises, the difference between an
image with and without threshold value is represented in
Fig. 3.

Fig. 1. MRI rendering with a 1D transfer function. In (a) skin has low
opacity and not shifted color value. In (b) all colors have been divided by
its opacity.

One of the practical questions here is how to

calculate the start and the end positions of the ray. One of
the simplest ways is to render a cube in additional textures
with textures coordinates on it with FRONT and BACK cull
modes in the rasterization state, like in Fig. 2. Note that each
polygon has a front and backside, depending on the vertex
traversal order. Having that, ray positions might be sampled
from the textures. The other way is to calculate ray Fig. 3. simple MRI rendering with shading (a) with a threshold value and
directions and cast them from the camera. It can be (b) without threshold value
completed in the local MRI texture space directly. But in
this case, it’s necessary to find a collision between ray and Transfer function
[0,1] XYZ coordinates, because calculations in exterior of
The most important side of MRI visualization is
this cube are excess. Both approaches are applicable and any
how to make different tissues distinguishable. MRI scanners
can be chosen.
provide information about hydrogen response at the point
and different tissues have various hydrogen concentrations,

37
the simplest way is to assign different colors and We can use a 2D or 3D transfer function, based on
transparencies to different densities. A look-up table is used scalar data, first and second derivatives along the gradient
for that. In DirectX and other graphic API, the look-up table [3, 4]. 2D histogram includes two axes, the first is the scalar
usually is a texture, which stores, in this particular case, data, the second is the first derivative along with gradient
optical parameters in pixels and scalar data along the axis. magnitude and red intensity is a number of combinations.
By this, we can assign optical properties to the field of MRI The 3D histogram, in addition, includes a second derivative
data directly and render it. Note that for sampling from an along the gradient. Example of the 2D histogram, based on
MRI texture it can be important to turn off interpolation artificial data, is presented on Fig. 5 (b). As can be seen, the
between voxels because interpolation between max value boundaries are represented by a few solo points. It happened
and zero can return randomly optical properties. due to the digital nature of using data (Fig. 5(a)).
The way described above is good. Nonetheless, this For better results, there is a need to apply Gaussian
way isn’t the best due to disturbances in MRI data. Another blur to an MRI texture. The result is shown in the next
way is to use methods from computer vision and underline figures: in Fig. 6(b) boundaries are represented by arcs, in
only boundaries between materials [3]. It increases the Fig. 6(c) as waves. In Fig. 7 2D histogram based on a real
quality of rendering because tissues are described by a MRI texture is represented.
single scalar value, but boundaries are described by a set of
two scalar values and it provides more combinations,
allowing more detail to be emphasized. It also increases
performance because modern shader models or modern
GPUs can pass calculations in if-body operators for empty
voxels in a few cases. Besides, it allows you to take the next
step in transfer-function generation, by expanding a transfer
function dimensional.
Boundaries can be detected by calculating the first
derivative along the gradient of scalar data. The example of
this is presented in figure 3. As can be seen, boundaries are
distinguishable.

Fig. 6. (a) is blurred by Gaussian with sigma = 3 and core size = 7 artificial
volume texture. (b) and (c) are the histograms of this texture

The first derivative along the gradient is the

magnitude of the gradient and it can be obtained with (3).
Fig. 4. (a) is a slice of MRI texture data. (b) represent the first derivative
magnitude along the gradient.
The second derivative along the gradient can be obtained
with Hessian [5] (3).

(5)

Using these, 2D or 3D histograms transfer function

can be generated. He [5] used a genetic algorithm to breed
transfer function. The system generated a few transfer
functions randomly and he chose the best, “breeding”
transfer function. Kniss described approach, where the user
selected the transfer function manually [4]. Also, neural
networks can be useful for finding a solution to increase the
quality of MRI textures [7, 9]. But in this way, it's important
Fig. 5. (a) is an artificial texture and (b) is a 2D histogram of this texture. to remember that neural networks are black boxes and they
Red points represent all combinations of scalar data and first derivative can hide important parts of the MRI texture or create
magnitude. artificial structures

38
 Where C is an output color, C0 is an initial color, ex is an
exposition.

Fig. 7. 2D histogram based on a real MRI texture.

IV. EEG SENSORS PROJECTION

For projecting data from EEG sensors to the brain,
the method like in deferred shading can be used [6]. The Fig. 9. The histogram, based on weak MRI texture.
positions of the brain surface, represented by boundary, are
stored into an additional texture (Fig. 8). With this, the
distance between the surface to EEG sensors can be
calculated. Having that, intensity can be represented as a
function from the distance between the sensor and the tissue
voxel.
There are a few methods of how to image intensity
to the brain. The first is to render intensity in additional
texture and blend it with the main image. Unfortunately, it
will not make good sense. The second is to create an
additional 3D texture, which will store intensity information
in the whole 3D MRI field. But the last way will decrease
performance.

Fig. 10. (a) constant transfer function. (b) linear transfer function. (c) linear
transfer function after compression

REFERENCES
[1] Lars G. Hanson. Introduction to Magnetic Resonance Imaging
Techniques. Danish Research Centre for Magnetic Resonance
(DRCMR), Copenhagen University Hospital Hvidovre. 2009.
[2] Milan Ikits, Joe Kniss, Aaron Lefohn, Charles Hansen. GPU Gems.
2007.
[3] Gordon Kindlman and James W. Durkin. Semi-Automatic Generation
Fig. 8. Positions of bone voxels in the local space of MRI texture. XYZ of Transfer Functions for Direct Volume Rendering. In IEEE
Symposium On Volume Visualization, pages 79-86, 1998.
coordinates are displayed in RGB.
[4] Joe Kniss and Gordon Kindlman. Multidimensional Transfer
Function for Volume Rendering. In IEEE Transaction on
Visualization and Computer Graphics, August 2002.
V. FUTURE WORK [5] Taosong He, Lichan Hong, Arie Kaufman, and Hanspeter Pfister.
Generation of Transfer Functions with Stochastic Search Techniques.
One of the problems of generating transfer In Proceedings Visualization ’96, pages 227–234, 1996.
[6] Jonathan Thaler. Deferred Rendering. TU Wein.
functions is the fact that scalar data depend on radio wave [7] Salman UI Hassan Dar, Muzaffer Ozbey, Ahmet Burak Catli, Tolga
response. It doesn't distinguish all tissue and the result of Cukur. A transfer-Learning Approach for Accelerated MRI using
Deep Neural Networks.
visualization very relies on the parameters of the MRI [8] Ljung P., Krueger J., Groeller E., Hadwiger M., Hansen C.,
process and disturbances. Ynnerman A.. State of the Art in Transfer Function for Direct
Volume Rendering. Computer graphic forum, 35(3), 669-691, June
The example of bad MRI data is represented by the 2016
2D histogram in Fig. 9. There are few arcs, which represent [9] Jia-Wan Zhang and Ji-Zhou Sun, Adaptive transfer function design
for volume rendering by using a general regression neural network.
the boundary between air and all tissues, which have the Proceedings of the 2003 International Conference on Machine
same scalar value. Learning and Cybernetics, 2003.
[10] J. Hesser, R. Männer, D. F. Braus, G. Ende & F. A. Henn. Real-time
The example of a problem with weakly direct volume rendering in functional magnetic resonance imaging.
distinguishable tissues is shown in Fig. 10. All tissues have Magnetic Resonance Materials in Physics, Biology and Medicine
volume 5, 1997
similar scalar value and it became identify only after tonal
compression (5)

(5)

39
 Analyses on Physics Textbooks for Secondary
Schools

Munkhbold Mangalsuren Purevdorj Dondovsambuu Sodov Tsedendamba

Department of Physics Department of Physics Physics Education Journal
School of Mathematics and Natural Science School of Mathematics and Natural Science Ulaanbaatar, Mongolia
Mongolian State University of Education Mongolian State University of Education [email protected]
Ulaanbaatar, Mongolia Ulaanbaatar, Mongolia
[email protected] [email protected]

Abstract - Textbook is committed to implement educational x It has to be attractive

standards and training program, to provide information for In the textbook for secondary schools, sentences have to
students, to let them acquire independent learning methods, meet the following requirements: it does not have two
motivate and activate them.
possible meanings, it must be clear, interesting and exciting
phrases and expressions, to provide as many tables,
Keywords—knowledge, measurement, regularity, conception graphics, charts, drawings, pictures, schemes and examples
as possible; it stipulates in which scopes the applicable laws
I. INTRODUCTION and regularity can be used without providing subjects and
matters that have not been proven scientifically and
Within a scope of definite topics, textbooks for
practically; to type learning formulae and measuring
secondary schools refer to principal materials of training
definitions with another colors or ciphers; to take examples
composing of texts, pictures, tables, problems for
from known and popular things for students; letter fond
confirmation, questions, tests and tasks on which basic
must be readable and likeable; and it is available to compare
knowledge for schoolchildren are clearly written. We would
the telling and explaining objects, things and subjects with
like to share my views with readers about how to make
another things/14,16,18,20/.
analysis on physics textbooks for secondary schools,
Teaching and pedagogical materials for pupils and
training programs and standards were changed, developed,
students must comply with the following requirements/15/:
and how to enhance them in the future. For doing so, we
x Its content must be transparent
have definitely concerned the correspondence of the content
sustainability and succession with training standards and x Ideal viewing size should be sufficient
composition by mainly focusing on regularity, idea, x Color must be suitable
dimension, level and practical scope from regularity, x Full use of the page area
conception, measurement of physical science that were x Not to fill readable area with supplementary
chosen as study materials. materials
Since victory of the People’s Revolution in x The dashed lines are clear and distinctive
Mongolia (in 1921) translated versions of the Russian x Writing as little as possible.
textbooks were used. The first ones that had been translated Since 1951 Textbooks “Courses of Physics” by authors
by D. Budari and published in 1929 and 1932 were books A.V.Pyerishkin, G.I.Faleev, V.V. Krauklis. I.I.Sokolov that
named after “Elementary textbook of physics” by A. V. had been used for the grades VI-X at secondary schools in
Tsinger with 2 volumes, 10 chapters and 296 subchapters. the Soviet Union were translated by N. Elbeg, M.
Textbook refers to clear meaning of training policy. Chimiddorej, D. Tserendagav, B. Lkhamsuren, B. Baldoo,
It is specific teaching materials containing general goals, E. Radnaa and N. Sodnom, published and used. Then in the
content, viewpoints, method and logics of the said science, 60s and 70s of last century, National Institute of Pedagogy
appropriating to the age and mental characteristics of the developed contents of textbooks with local authors,
student. Textbook is committed to implement educational published under Order by Ministry of Enlightment and
standards and training program, to provide information for applied them for teaching physics. So content, requirement
students, to let them acquire independent learning methods, and methodology of the Soviet Union were followed at that
motivate and activate them. Compared to scientific time.
simplified literatures, textbook is different by teaching Furthermore, it renewed educational program of physics
definite knowledge content based on facts and evidences in 1955, 1963, 1964, 1972, 1984, 1999 and 2014. Although
and expressing content with precedent knowledge and the training content of physics is applied with textbook,
continuity; it doesn’t cite any views or ideas of authors like workbook of problems, experience and lab works for show,
science books; pictures, decorations and independent supplementary materials of independent learning and studies
exercises must be obligatorily included; its composition and for students (elective subjects, preparation for Olympiads or
stylistics must absolutely be clear. Textbooks should enrollment exams at colleges, universities), teaching and
comply with the following requirements of scientific pedagogical manuals and supplementary materials for
theories, concepts and regularity system: teachers; we have made analysis only on the textbooks that
x It must, scientifically, be honest were approved officially.
x Its terms should be rational
x Language and composition must be clear
x Classify major and minor matters

40
 II. MAJOR PART After having analyzed two textbooks together at level
This time we have analyzed the named textbooks of of basic education, we found out that training of physics was
physics for secondary schools composed by national conducted at two stages by imitating foreign country. We
authors, having definitely considered the compliance with consider it would be better, much clear and accessible if
program and standards, compatibility to students’ interests, fundamental regulations, laws and facts of physics are
needs and feature of training environment, design, sufficiently studied in high schools; and integrated
pedagogical solution, optimal application of teaching knowledge of natural sciences are taught in the grades VI
methods, aids and tools, light and hard exercises and tasks and VII with physical and other natural phenomena.
for independent accomplishment by students. Then we have PHYSICS-VII. Physical basis of mechanics was clearly
divided textbooks of physics into four main categories in the provided according to the program of subject physics at that
followings: period. Besides that, it is appreciated that the most important
A. Textbooks of physics in 1961-1979 things for each chapter were taught in manner of
B. Textbooks of physics in 1984-1986 conclusions. But there were too many formulae with and
C. Textbooks of physics in 1997-2001 without extraction in sections §12, 22, 42, 43, 50, 52, 53, 64,
D. Textbook of physics in 2014 that was prepared and 69. It seems that §5-7 and §41-43 can be integrated.
under core program. PHYSICS-IX. There was a preface for beginning parts
I. PHYSICS–VII. Each topic of this textbook has of this textbook but there was no introduction for ending
tests, pictures, explanation of terms, Mongolian names and parts. It seems that its editors might not review some
nomenclatures and is in good order. For instance, it subjects of this textbook. Conceptual theory was wrongly
mentioned about solar energy and jet engine. But tables of explained in §18 and §24; the level of material science on
physical constants were not numbered. electricity in vacuum and semi-conductor in chapter X
PHYSICS-X. Besides aiming at showing physical “Electricity in various environments” is too weak.
fundamentals of any phenomenon and developing Moreover, volume of the textbook became thick since some
independent skills of students, it sets goal to show scientific topics (§31, 32, 33) were written too detailed. So it may
achievements and discoveries at that periods. Thus, for the integrate the too-detailed things and make conclusion, then
first time the chapter “Fundamentals of Relativity Theory” make up other concepts, content and structure in manner of
was provided in this textbook and prepared in excellent interpreting laws.
order. But it observed that order of group topics was lost. PHYSICS-X. The content of this textbook fully
PHYSICS-VIII. It is consisted of two main parts: complied with the program and it might adhere an ideology
mechanics (1-11 chapters) and molecular physics (12-16 of the Russian Federation too much. It didn’t apply the
chapters). It clearly explained theoretical materials with opportunity to densify energy of mechanical and
experiences, practices, pictures and graphs, made derivation electromagnetic wave in huge amount; spectrum was
of formulae and mentioned in details about how to apply classified and some of them were explained in details even
them in practices. For instances, it became advantageous by there is no reference of spectrum (chapter 7). In our views, it
teaching relativity of movement, equations of smooth may be optimal and clear if this chapter would have been
acceleration and deceleration, body weight, weight gain, written with method of deduction.
weightlessness, centrifugal force, body movement on a Questions conform to texts in textbooks for the 7 th
sloping plane, conservation law in mechanics, work and grade. 20% of questions could be answered after getting
energy with tests and formulae, and resolving relevant information from other books, teacher or taking additional
problems. information. Its level and content are conformed, it is
PHYSICS-IX. The crew composing of 5 members difficult for children to understand hydraulic equipment,
such as lecturers and researchers from Mongolian State pressure measuring device and pumps, there is no reference
University of Education, National University of Mongolia on mechanism. Although mechanic movement is written
and Institute of Pedagogy wrote it. Its content was there, thermal movement is not provided. So it bought the
comprised of theory, methodology to solve problem, tasks reason that the textbook of physics for the 7 th grade is too
and exercises, lab works; its terms were clear and smooth, thick. Some teachers couldn’t teach the content in time.
mostly it paid more attention to explain theories by Since numbers of some pictures are not mentioned in texts,
experiences and practices, describe by pictures and graphs, they are incomprehensible for students; principle of colorful
and to apply in practices. Additionally, it provided new textbook was broken even though theoretical and practical
chapters such as magnetic properties of matters, alternative comparison was chosen well. However, there are many
current. But some pictures were not clear as follows: picture things to show and design are good, pictures are placed in
84 – aluminum purification furnace, picture 115 – deviation right location, questions and exercises are optimal for
of a set of electronics in magnetic field, picture 182 – students.
scheme for long-distance transmission of electricity, picture In the textbook for the 8th grade, many questions on
182 - thermal phenomena requires students to read other sources
PHYSICS-VII. The most contents of this textbook and books and to have logical thoughts. There are not few
(topics or articles 1, 3, 10, 16, 32, 33, 56, 65) coincided with questions to answer by reading textbooks of high school.
briefed contents of corresponding topics in textbooks for the This textbook could respond to questions about
grades IX and X; its concepts were not completely electromagnetic, light, astronomic topics. Materials for the
identified; main values were bypassed and explained in poor 7th grade were written on the first three chapters. Standards
orders. We don’t agree such manner of educating of electromagnetic ad light content meet appropriate
knowledge of physics with periodical coincidence. Many requirements. But the introduction about properties of the
topics were not clear for the 7th grade students and it didn’t sun and stars in astronomy exceeded from appropriate
express physical value of the content; and much standards. Design, composition, question, test and tasks
simplification might give false concepts for students. were included appropriately. In picture 30, lessons such as

41
temperature at the bottom of lake in summer and winter, Mechanic movement and Mechanic energy, Electricity and
magnetization by influence, projection of celestial bodies in Magnetics, Energy supply” in the textbook for the 8 th grade
the sky are difficult for students, especially in picture 10.4, have completely met the relevant requirements. Except that,
students are not able to work on it by looking at its annex. tasks designed for students’ knowledge and to resolve
The design of the textbook is excellent, especially its problems have been optimal. Instead of stating that current
background color of table and graphs are much suitable, run on diode, engine and bulb on the pages 88 and 89, it
many real pictures make it believable. Unfortunately, some would be more understandable to state that current would
mistakes were found in it as editors didn’t review, detect run through diode, engine and bulb. It needs to rectify
and rectify mistakes. percentage of heat loss more than 100% in the picture on the
In the textbook for the 9th grade, there are many tasks page 142.
for students to make conclusion and answer after making Since the physics textbook for the 9th grade is the last
discussion even too little information. Teachers don’t textbook for students to acquire basic education of physics,
believe that the 9th grade students would answer 128 it is appreciable that authors focused on giving orderly
questions (which occupy 30% of the textbook). In order to knowledge and concepts to students. To do that, it was
answer these questions, they need to read and know many written well for students to have universal concepts, skills
things. How many such kind of materials do we have? In and knowledge about thermal movement, energy, magnetic
picture 198, it didn’t mention with which kinds of materials and electric fields, electric circuit and light based on
glass and plastic sticks were rubbed; it put electroscope accumulated experiences and knowledge in the 7th and 8th
exclamation mark was put in picture 195; in pictures 272- grades. In order to express these concepts, authors have
276 exactly what materials of semi-conductor were optimally chosen pictures in texts.
described in text; it didn’t mention what was marked by the The 10th grade textbook for secondary education is
circle with arrow in picture 291; it didn’t concern about how consisted of main chapters such as mechanics, heat, wave,
to use picture 321. Authors didn’t predict and imagine how sound, electromagnetism, problems for processing
students could answer such questions. Furthermore, editors measurement results and letting students solving those
didn’t observe that pictures 233 and 243 were wrong. problems. It was written good in manner that students could
Unfortunately, it was published that way. Although the enrich and study matters such as description of body
content of this physics textbook meets standards and levels, position, displacement, movement and forces by vectors;
several things such as Cavendish experiment one of main description of weight and center of gravity by torque /
experiment, experiment to determine coefficient of friction, moments of force; use of thermal process phenomena in
electric statics coincided with the 7th and 8th grade practices; wave and sound by expressions of wave
textbooks. Choice of pictures is good and composition is refractions; electromagnetic field by modern electronic
affordable. But we would like to say it is extra vintage that system, circuit, elements and logic functions respectively
there is each title of chapter 23 and phrase that “this chapter based on the acquired knowledge and competence about
shall study those matters”. mechanics at level of basic education. It is very appreciable
It is impossible to answer question after student reads that another feature of this textbook is to make definite
information in the textbook for the 10th grade. It is similar to measurement on precedent theoretical knowledge, to make
the 9th grade’s. For example, pictures 39, 40, 41 describing conclusion of its results based on digital processing, and to
rule for establishment of body shapes in mirror and lenses, solve problems (on pages 126-148) relating to the said
interference, interpretation about electromagnetic waves by topics. The content and level of these problems have met
generation and distribution, picture 113 for polarization of concepts and conclusions of texts.
the light. Its content and level complied with appropriate The 11th grade textbook stating that the movement
standards. Students don’t have any chances to practice using quantity and impulse in the chapter “mechanics” by
picture of sky with stars on the pages 125-126 in chapter 13 Bernoulli’s equation; fundamentals of ideal gas equation
and characteristics of some sharp stars. Even though design and molecular genetic theory in the chapter “Molecular
and composition are generally affordable, first six chapters physics” were enriched by the second law of
have similar conceptual design. Design is good, but we thermodynamics and its applications respectively; was
deliberately note that pictures 116, 118, 153, 164, 184 and written understandable for students. The topics “Electric and
205 should be colored for pedagogical requirements. Magnetic fields” were mostly interpreted and explained by
The students who have taken information from the pictures without hard formulae. It was excitingly explained
textbook for the 7th grade written under new program can about electricity in semi-conductor and Ohm’s law for
answer questions and carry out tasks after reading it well. complete circuit based on applicable examples. Also this
But at the beginning of chapter 1 “general properties of textbook has excellent attractive experiences and
body”, it writes about how to measure length, weight and experiments. For instance, much clear expression of
density of the body without explaining the term “body”. Its mechanic and harmonic fluctuations, interactions of their
content duly complied with the program. The chapter parameters; shutdown and forced fluctuations, and specific
“Astronomy” is the most difficult part for students to read explanation about transverse wave propagation mechanism
punctually and work hard. Some concepts don’t conform to based on Huygens-Fresnel principle. It is understandable
titles and preface. For instance, current and electric that changes in frequency on the Doppler effect were
magnetic properties are not included in texts. Since it didn’t described by picture. It is available for teachers to use it for
clearly distinguish the difference between two concepts teaching other waves.
“magnet” and “magnetic”, page 17 was explained by spring Nowadays, in order to make contents understandable
scale and three rings on the page 60 interpreted with word and accessible for students except than main content of the
“five”. program, it created electronic textbook with CVDs / CDs by
We would like deliberately note that the contents and entering data, establishing diagram, scheme, graphic
program in four chapters “Forces and Interactions, descriptions, with studying materials for micro lab works,

42
modern and future technology, information receivable and letting teachers sharing and discussing their views. We
affordable film by answering questions of students, real trial considered that, as results of such trial, the content of
and experiment, games, brief history, interviews, video physics textbook and all subjects would be stabilized and
recording and discussion, electronic tasks, assessment qualified, it would make strong contribution to development
mission and tests. It requires to creatively and effectively of general education system in our country by determining
apply them for training of physics. demands of the subjects (student, author, teacher, policy
The issue for the training of physics that is waiting for maker / developer, methodologist, researcher, employer,
solution is how to completely and optimally insert problems professional association, unions) that involve in educational
in the textbooks for students of that class. Since such issue relations.
have not been fully resolved till today, it requires some 6. If electronic textbook is invented and distributed
authors to include problems at the end of each chapter / teachers and disciples at all stages of schools, we may not
topic, every teacher to choose problems based on his / her reach the desired approaches but we could achieve certain
own methods and experiences. Finally, it needs to create, results. The experiences of electronic training during the
compose and publish new core program, content, structure outbreak of pandemic disease Covid-19 has showed and
and directions of physics textbooks for “Compilation of warned us it is time for us to get to such work. So it needs to
Physics Problems” at the levels of basic and secondary restore and renew the precedent experience (composition
education. and regulation for such non-staff council have been ratified
under order No.163 by chairman of State Committee of
III. CONCLUSION AND POINT OF VIEW Science, Technique and Higher Education from Jun 10,
As for 18 physics textbooks for general education 1989; order No.426 by Minister of Education dated from
schools that we have studied, the textbooks that were used Nov 04, 1991; annex of order No. 03 dated from Jan 05,
by us till 1951 were directly translated from textbooks of the 1993) to establish Training Methodology Classified
Soviet Union. As the program belonged to the Soviet Union, Councils with authority to make assessment and publish
its content had the same meaning. Since 1951, physics training program and textbooks of general education schools
textbooks for the 6-10th grades used in the Soviet Union and to employ chairman, deputy chairman, and secretary
were translated and applied under order by Minister of under order by minister. Finally, we consider that the said
Enlightment of Mongolia. measure would enhance quality of textbooks, level of
1. In the textbooks in 1961-1979 written by national training program, relationship between secondary, higher
authors, it paid more attention to provide polytechnic education and continuity of classes would be improved.
education for students based on definite goal of the country,
to introduce scientific inventions in production and practices REFERENCES
and to show their applications. [1] Ya. Lyernyera, N. M. Shakhmayeva, “What a textbook should be:
2. The textbooks written in 1984-1986 imitated didactic principles of construction, in sequence,” Part I-II, M, 1992.
foreign countries. During organizing and conducting [2] V. P. Bespalko, “Theory of a textbook: Didactic aspects,” M, 1988.
trainings of physics at two stages (incomplete and complete [3] A. Sokhor, “About the didactic processing of science materials in
textbooks (taking examples of physics), problems of school
secondary education), it has shortage in features of textbooks,” M., 1978.
Mongolian national thoughts, accumulated teaching
[4] J. Gombojav, S. Batkhuyag, J. Battseren, D.Ragchaa, “PHYSICS-
methods and experiences of teachers, introducing new VI,” Z. Badamsambuu and Ts. Nyamdorj, Eds. 1985.
technique and technological inventions, focused little on [5] L. Damba, Ts. Ragchaa, E. Radnaa E, “PHYSICS-VII,” S. Batkhuyag
integrating content of physics training. and J. Battseren, Eds. 1984.
3. The physics textbooks in 1997 – 2001 were written [6] Ts. Nyandag, Ochirbat, S. Doyoddorj, “PHYSICS-VIII,” G. Byambaa
according to educational standards of physics and and L. Damba, Eds. 1985.
astronomy for the first time. Their term and formulae were [7] N. Altangerel, G. Byambaa, D. Khaltar, “PHYSICS-IX,” D. Chultem
official, been taken into consideration of Mongolian and G. Gombojav, Eds. 1985.
household and living conditions fully suited for students, [8] D. Dambasuren, D. Gurdorj, N. Zundui, D. Tseren, “PHYSICS-X,” ),
contained traditional and new content of astronomy. D. Chultem and Z. Badamsambuu, Eds. 1986.
4. Physics textbook of basic education according to [9] B. Burmaa, M. Ganbat, B. Jadambaa, “PHYSICS-VII,” S. Batkhuyag
newly approved core program are advantageous than and S. Danzan, Eds.1997.
previous old textbooks that their design and expressing [10] M. Ganbat, B. Burmaa, D. Batbold, R. Maabazar, “PHYSICS-VIII,”
M. Ganbat, Ed. 1999.
methods are suitable for age and mental characteristics of
students. After having studied these, we have made [11] G. Gombojav, O. Lkhagva, G. Batbyamba, “PHYSICS-IX,” B.
Jadambaa and G. Bold, Eds. 2001.
conclusion that ground to study physics at level of complete
[12] G. Gombojav, O. Lkhagva, G. Batbyamba, “PHYSICS-X,” B.
secondary education was set up. The physics textbooks for Jadambaa and R. Maabazar, Eds. 2001.
the 10th and 11th grades have carried out its obligations to [13] M.Gongor, Ts. Sodov, Kh. Khovd-Aba, N. Tugjsuren, Ts.
provide systematic knowledge to students at high schools Erdenechuluun, “About textbooks and manuals which have been used
and to make them be interesting in physics. for training of physics,” Professional Education, vol. 10, pp 76-81,
5. Since we have had standards of basic and 1992.
secondary education, it would be better to compose and [14] N. Khajidsuren, Ts. Sodov, “Textbook – important pedagogical tools,
publish textbooks at such two levels. That would bring us scientific materials,” vol. 01, pp. 87-93, 1998.
opportunities to get away from the economic pressure in [15] Ts. Sodov, J. Baigalmaa, “Technology to develop students' general
and experimental learning skills in physical education,” Education
manner that contents of textbooks would not be changed Research, vol. 3, pp 30-36, 2004.
every year, not publishing each year, uncertainty of content
[16] A.V. Khutorskii, “Place of the textbook in the didactic system,”
for teachers, students and parents, to eliminate delayed Pedagogy, vol. 4, pp. 10-18, 2005.
delivery of textbook, to enhance sufficiency by completely [17] “Mongolian Agency of Standard and Metrology (MASM): Physics
testing results of contents in the published textbooks and by and Astronomy in basic and secondary education standards,” 1998.

43
[18] I. Tuya, “ Textbook analysis modeling,” Education research, vol. [32] E. Damdinsuren, S. Batkhuyag, Ts. Sodov, “PHYSICS,” A.Samdan
4(50), pp. 31-35, 2009. and P. Zuzaan, Eds. 1992.
[19] UNESCO “Guidebook on textbook research and textbook revision”, [33] E. Damdinsuren, S. Batkhuyag, Ts. Sodov, “PHYSICS,” A.Samdan
guideline for enhancing quality education, 2010. and P. Zuzaan, Eds. 2013.
[20] J. Narantuya, “Issue for assessing quality of secondary education [34] R. Jigjidsuren, D. Khaltar, J. Gombojav, “PHYSICS-VII,” D.
textbook,” Studies on education, vol. 9, pp. 51-57, 2011. Ragchaa, Ed. 1963.
[21] J. Narantuya, “Textbook: What a dream should be?” Education [35] G. Byambaa,G. Dodon, Ts. Nyandag, “PHYSICS-VIII,” G. Dodon,
research , vol. 4, pp. 31-36,2012. Ed. 1975.
[22] B. Bayarmaa, “Textbook development: at level of some foreign [36] G. Byambaa, D. Gurdorj, D. Tseren, “PHYSICS-X,” G. Dodon, Ed.
countries,” Education Research, vol. 10(98), pp. 17-22, 2013, 1975.
[23] R. Jigjiddorj, “Pedagogical issues of physics”, 1996. [37] G. Byambaa, G. Dodon, D. Khaltar, Ts. Ragchaa, Ts. Nyandag
[24] M. Nasanbuyan, G.S. Manyeev, “Some issues in teaching of physics “PHYSICS-IX, ” G. Dodon, Ed. 1975.
at secondary schools,” Gombojav , Ed. 1962. [38] M. Ganbat, B. Burmaa, Ts. Amartaivan, J. Dulguun, Ya.
[25] Ts. Sodov, M. Gongor, L. Udval, “Fundamentals of physics Munkhsaikhan, S. Gendenjamts, N. Sainbayar, A. Buyandelger,
teaching,” B. Jadambaa and Kh. Dondog, Eds. 2018. “PHYSICS-VII,” N. Tugsjsuren, Ed. 2019.
[26] B. Purev, B. Orgoi, “MECHANICS-VIII, Learning of physics as an [39] M. Ganbat, B. Burmaa, B. Badam, J. Dulguun, Ya. Munkhsaikhan, S.
elective subject,” 1979. Gendenjamts, B. Purevjalm, “PHYSICS-VIII,” N. Altangerel, Ed.
2019.
[27] S. Batkhuyag, Ts. Sodov, L. Jamts, “ Learning of physics as an
elective subject: grades IX-X,” N. Altangerel, Ed. [40] M. Ganbat, Ts. Amartaivan, Kh. Batbold, J. Dulguun, Ya.
Munkhsaikha, B. Purevjal, “PHYSICS-IX,” E. Damdinsuren, Ed.
[28] U. Bolormaa, Kh. Dondog, Yo. Jantsan, R. Maabazar, D. Oyun, D. 2019.
Oyunchimeg, Ts. Sodov, “PHYSICS-ASTRONOMY (dedicated for
basic education learners),” Ts. Sodov, Ed. 2004. [41] M. Ganbat, B. Burmaa, Ts. Amartaivan, B. Bat-Otgon, G. Zorigt, Ya.
Munkhsaikhan, D. Uurintuya, B. Tuvdendorj, “ PHYSICS-X,” Sh.
[29] N. Amarjargal, J. Baigalmaa, I. Batmunkh, M. Gongor, R. Jargalsuren, Ed. 2019.
Darkhijav, Ch. Dashnyam, L. Dolgormaa, D. Demchigsuren, D.
Otgonbayar, B. Ulziikhutag, Ts. Sodov, N. Tugjsuren, B. [42] M. Ganbat, Sh. Chadraabal, O. Altangoo, Balt-Erdene, Bavuudorj, D.
Tsedenbal, L. Tsedendamba, “ PHYSICS-ASTRONOMY (dedicated Uurintuya, Ch. Gonchigdorj, “PHYSICS-XI,” Sh. Jargalsuren, Ed.
for high school pupils or pupils are expected to admit 2019.
colleges/universities),” Ts. Sodov, Ed .2004. [43] “PHYSICS – Core academic program of basic education (grades VI-
[30] E. Damdinsuren, S. Batkhuyag, Ts. Sodov, “PHYSICS,” A.Samdan IX) compiled by Ministry of Education, Science and Sports”, 2015.
and P. Zuzaan, Eds. 1977. [44] “PHYSICS – Core academic program of secondary education (grades
[31] E. Damdinsuren, S. Batkhuyag, Ts. Sodov, “PHYSICS,” A.Samdan X-XII) compiled by Ministry of Education, Science and Sports”,
and P. Zuzaan, Eds. 1985. 2016.

44
 Autonomous Unmanned Aerial Vehicle
Development: MAVLink Abstraction Layer
Marco Stephan Batbayar Battseren Uranchimeg Tudevdagva
Professorship of Computer Engineering Professorship of Computer Engineering Professorship of Computer Engineering
Chemnitz University of Technology Chemnitz University of Technology Chemnitz University of Technology
Chemnitz, Germany Chemnitz, Germany Chemnitz, Germany
[email protected] [email protected] [email protected]
chemnitz.de

Abstract—In the domain of Unmanned Aerial Vehicle modular software components running on the on-board
(UAV), the autonomy is getting more important. In order to computer to execute the autonomous mission. Some of these
reach a higher autonomy level, certain key software components components are the Camera Gimbal Controller (GCC), which
have to be implemented. One of that is a middle layer, which lies controls the UAV’s gimbal, the Feature Point Detection (FPD)
between the decision-making higher-level and hardware- which detects objects that are to be inspected, as well as the
controlling lower-level components. This study aims to abstract Expert System (EXS) which chooses appropriate flight paths
the UAV control process by conceptualizing an abstraction layer and makes real-time decisions. A separate component, the
for the UAV communication protocol. Control Handler (CTH), manages the communication and data
conversion between the GCG, the FPD, and the EXS. In
In the context of the APOLI project, a MAVLink
Abstraction Layer has been implemented to control and
addition to the ODROID-XU4, a second board, the so-called
communicate with the UAV flight controller. Our proposed Flight Controller (FLC), controls the individual actuators and
solution simplifies the usage of the MAVLink protocol sensors of the UAV on a lower level.
significantly for the higher-level components. Also, the result The protocol used for the communication between the
indicates that an abstraction layer does have an advantage over FLC and the ODROID-XU4 is Micro Air Vehicle Link, or
existing APIs. MAVLink [5]. As this protocol is too complex to be handled
by the Control Handler itself, and the existing MAVLink APIs
Keywords—unmanned aerial vehicle, UAV, autonomous
mission, abstraction layer, MAVLink protocol
are not compatible and not supporting our system, the
MAVLink Abstraction Layer (MAL) software component is
I. INTRODUCTION proposed. The reason why these existing APIs are not the best
solution is explained in detail in the Literature Review
Automated Power Line Inspection [1]-[3], also referred to
chapter.
as APOLI, is a research project carried out by the
Professorship of Computer Engineering at the Chemnitz The MAL abstracts the MAVLink protocol to be usable on
University of Technology. It aims to develop an Unmanned a higher level and allows the companion computer to issue the
Aerial System (UAS) able to autonomously inspect power flight commands on a high abstraction level. Furthermore,
lines and detect damages of them. In the context of this requested UAV data, e.g. GPS position or flight mode, can
project, we are developing an Adaptive Research Multicopter also be provided in a ready-to-use format.
Platform (AREIOM) [2], which is a modular layered
architecture, that can be used in any autonomous UAS. II. RESEARCH QUESTION AND REQUIRMENTS
In this study, we will answer the question of how to
abstract the MAVLink protocol for an autonomous UAS,
considering the given requirements. Our main objective is to
separate the decision-making layer of the autonomous system
from the hardware or low-level software layers by
implementing a MAVLink abstraction API.
The requirements are, firstly, and most importantly, it
must be able to communicate with the FLC via a serial port so
it can send data to and receive data from the Flight Controller.
Additional communication interfaces that are to be supported
for the communication between the MAL and the FLC are
UDP as well as TCP sockets. These socket connections are
necessary for integration into a software-in-the-loop
simulation in which the FLC is running as a software
component.
Fig. 1. APOLI Software Architecture [2].
Furthermore, the MAL must provide a communication
In our autonomous system, the UAS is controlled by a interface for the CTH-MAL communication. However, at the
single on-board computer, namely an ODROID-XU4 [4], time of this article’s publication, the CTH has not been fully
being directly mounted on the Unmanned Aerial Vehicle developed yet. Therefore, the MAL also mimics some basic
(UAV). As illustrated in Figure 1, there are a number of logic of the CTH itself and provides a connection with the

45
Expert System. Communication with the Expert System is UAV is armed using a loop or ignore a potential failure of this
realized via file I/O. To issue a flight command, the software command altogether.
component, i.e. CTH or EXS, writes an encoding of the
command into the corresponding connection. The MAL then IV. IMPLEMENTATION
needs to read this command, decodes it, and eventually As declared before, the MAL should be implemented
executes it. If there is any response to this command provided using a compiled programming language to be more efficient
by the UAV, e.g. the GPS position, it may then be written back in comparison to an interpreted language. For this, the C
into the connection by the MAL. After this, the CTH or EXS language is feasible, because of the availability of the
will read and evaluate this response. MAVLink C library. To further improve maintainability and
In addition to these functional requirements, the MAL code readability, the MAL is implemented in C++.
must also be efficient so UAV commands are executed as soon TestVehicle CTHVehicle EXSVehicle
as they are issued. Furthermore, the MAL also has to fulfill
high-reliability requirements. A failure of this component, Vehicle
while the UAV is flying, is disastrous and may result in +VehicleRoutine()

crashes of the vehicle. Last but not least, the MAL must
provide a cross-platform solution. While the APOLI’s MavlinkAbstraction
1 1
C2Handler
-currentRequest : Request
+Arm(ResponseArm*)
ODROID-XU4 runs on a Linux distribution, the simulation …
1
+HandleRequest(Request*, Response*)
loop runs on a Windows machine. +TakeOff(ResponseTakeOff*) +OnMAVLinkMessageReceived(mavlink_ message_t*)

1
MAVLinkConnection 1 1 MAVLinkConnectionReadingThread
III. LITERATURE REVIEW #connect()
Currently, there are several MAVLink APIs available. The #onMessageReceived()
#disconnect()
1 1 MAVLinkConnectionHandlingThread

most popular MAVLink abstraction API is DroneKit [6]. It +OnMAVLinkMessageReceived(mavlink_ message_t*) 1 1 MAVLinkConnectionWritingThread
+SendMAVLinkMessage(mavlink_message_t*)
provides interfaces to control any UAV using high-level
functions. Using this, flight commands like arming or a take- Request Response
off are easily issued and handled by the DroneKit API itself. +Execute(Response*, …)
+OnMAVLinkMessageReceived(mavlink_ message_t*)
However, this API is not feasible for use in the MAL of the
ResponseArm
APOLI project. This is because DroneKit only provides an RequestArm +success : bool
API for the Python programming language [7]. As Python is ⋮ ⋮
an interpreted language, the interpretation of the MAL itself RequestTakeOff ResponseTakeOff
requires additional resources that are to be avoided. Instead, a -altitude : float +success : bool

compiled programming language is preferred.

Fig. 2. Simplified MAL Class Diagram.
Another MAVLink API is MAVROS [8]. It is a package
developed for the Robot Operation System (ROS) framework. A. MAVLink Abstraction
MAVROS also provides, just like DroneKit, interfaces As shown in Figure 2, an abstraction of the MAVLink
abstracting MAVLink commands. Unlike DroneKit, however, protocol is implemented in the MavlinkAbstraction
MAVROS can be used in the C++ programming language,
class. This class provides easy-to-use functions for the flight
and therefore, it can be integrated in a compiled language.
commands, e.g. Arm() or TakeOff(). Upon calling one of
That being said, MAVROS also entails some drawbacks.
these abstraction functions, a command-specific so-called
Firstly, it is relatively limited in its functionality. For example,
MAVROS does not support issuing a reboot of the FLC. Request object is created. This object stores all parameters
Another disadvantage of the API is that officially it is only of the flight command, e.g. the take-off altitude. Additionally,
supported on Linux. it contains routines for the execution of the MAVLink
command, i.e. encoding it to a MAVLink byte stream and
A third MAVLink API can be obtained from the website sending the encoded MAVLink message. The Request
of MAVLink itself [9]. There, they provide a C library for the object will then be passed to the C2Handler. This class
MAVLink protocol. This library does not abstract the manages the communication to the Flight Controller on a
MAVLink commands themselves in any way but instead it higher level. This means that it executes requests, forwards
provides functions to encode command parameters to byte received MAVLink messages to them and also handles some
streams equal to the MAVLink representation of that basic MAVLink messages itself. For example, upon receiving
command. Furthermore, it provides functionality to decode a status message from the Flight Controller, the C2Handler
the byte stream of received MAVLink messages so they can outputs the text of the message to the console. By outsourcing
be evaluated as such. Additionally, the authors claim that this the individual requests to separate classes, instead of
C library is “highly optimized for resource-contained systems implementing all of them in the C2Handler itself, this class
with limited RAM”. For these reasons, this library is more and the requests themselves become much more manageable.
feasible for application in the MAL of the APOLI project. In fact, in an earlier implementation of the MAL, this
Furthermore, writing our own MAVLink Abstraction based separation did not exist which led to an increasing number of
on the MAVLink C library allows for more efficient use of the lines of code that became harder to maintain over time. As
processor. For example, if the arming of the UAV is to be elaborated before, the C2Handler forwards MAVLink
executed, our own MAL process can change into the blocked
messages to a Request object, if there is one currently
state and wait until an acknowledgment from the UAV has
been received. It can then evaluate this acknowledgment to executing. This is because a Request not only consists of
check whether the arming was successfully performed. Using sending a MAVLink command but most of the time also
DroneKit, we would either have to actively check whether the evaluates responses from the FLC. Such a response may be an
acknowledgment or, if requested, flight data of the UAV.

46
 The lower level implementation of the communication to EXS MAL FLC
the FLC is implemented in the MAVLinkConnection
class. This abstract class provides functions to send and Request Flight Data
receive MAVLink messages provided as the low-level Request Flight Data
message structure of the MAVLink C library. The abstract Send Flight Data
component of this class is the connection to the FLC itself.
Because this connection must be establishable via a serial port, Send Flight Data
a UDP socket, or a TCP socket, there are three child classes of Send Flight Command (if any)
MAVLinkConnection. Each of these classes overloads the Execute Flight Command
abstract function to open and close the connection. In addition
to this, MAVLinkConnection also consists of three
Fig. 4. Sequence Diagram showing the EXS-MAL-FLC Interaction.
threads.
x Reading Thread: Permanently reads bytes from the The second vehicle is the EXS Vehicle. This class is used
connection, builds a MAVLink message using these for communication with the Expert System. Figure 4 shows
read bytes, and add it to a queue. the communication sequence of this interaction. The Expert
System first requests the UAV’s flight data from the MAL
x Handling Thread: Unblocks, dequeues the received which it provides by using its abstraction functions. After the
MAVLink message, and evaluates it by executing the MAL received the flight data, it sends it to the EXS. Then, the
appropriate function of the C2Handler. EXS evaluates this flight data and, if appropriate, sends a
x Writing Thread: Writes queued MAVLink messages flight command to be executed to the MAL. This flight
to the connection. command is then executed by the MAL by using the
corresponding abstraction function. Currently, however, the
The Reading Thread exists because the reading from the EXS does not exist in a state in which it is useable in such a
connection is implemented as a synchronous operation. This manner and therefore the EXS Vehicle does not implement
means that reading blocks until data could be read. This is to this sequence. Instead, the EXS Vehicle first requests the
avoid busy-loops. Furthermore, writing to the connection may UAV’s flight data itself. Upon receiving it, it is being written
also block. This is especially true if a TCP socket is being into the input file used for the communication with the Expert
used. To avoid blocking while sending a MAVLink message System. The EXS Vehicle then executes the Expert System
on a higher level, they are therefore queued for sending and itself once which modifies the output file. This output file then
sent to the Writing Thread. Lastly, the handling of received contains rules on how to handle the UAV data, i.e. the flight
MAVLink messages is also outsourced to the Handling command to execute.
Thread. This is so the Reading Thread immediately returns to
reading from the connection. Therefore, the chance of losing The last vehicle is the CTH Vehicle. This vehicle provides
any data sent from the FLC is minimized. a basic loop intended to be used for the CTH communication.
For the connection with the CTH, shared memory is chosen as
B. Vehicles it provides the lowest overhead and arbitrary data access is
Figure 3 illustrates the architecture of the implemented easily performed. More specifically, it permanently reads the
MAL. In the architecture, the specific use cases of the MAL shared memory block’s command and, if there was one
are implemented in so-called Vehicle classes. Here, a provided, executes it. In the future, the MAL shall also write
vehicle denotes a specialization of the all of the UAV flight data into the shared memory.
MAVLinkAbstraction class itself. Currently, there are V. EVALUATION
three such classes.
The MAL has been tested in several environments. More
EXS
MAL UDP or specifically, the Test Vehicle has been used to evaluate the
File TCP or
EXS Vehicle MAVLinkAbstraction
Serial abstraction functions of the MavlinkAbstraction, e.g.
FLC
Shared Test Vehicle
C2Handler MAVLinkConnection
arming. These commands have been tested in three phases. As
CTH Memory
CTH Vehicle a first phase, they were being evaluated in the simulation loop
of the APOLI project. For this, the MAL has been built on a
Fig. 3. Simplified Architecture of the MAL . Windows machine. The MAL then has been started with
command-line arguments specifying the address and the port
Firstly, the Test Vehicle is primarily created for testing and of the simulated FLC (more precisely, these arguments are
development purposes. It takes input from the console and being passed to the MavlinkAbstraction class which
executes the given command. If appropriate, additional then creates an object of the appropriate sub-class of
parameters are also read. The execution of the command then, MAVLinkConnection). All fifteen abstraction functions
as described above, may block until the corresponding have been evaluated over a number of times. This includes the
response has been received. For example, there exists a check whether the actual sent byte stream represents the
command that retrieves all relevant UAV flight data such as correct command, which has been done by logging these
the GPS position. This command, which itself consists of streams into a file, as well as the evaluation of the received
multiple individual commands, blocks until all flight data is response data. These response data had to match their
received. That data is then output to the console. expected data in the corresponding flight environment. One
issue that came up in this evaluation step was that several
MAVLink commands were not supported by the FLC used.
After the corresponding abstraction functions have been
evaluated, the other two vehicles, i.e. the EXS and the CTH

47
Vehicle, have been evaluated as a whole. These tests VI. CONCLUSION AND FUTURE SCOPE
succeeded, apart from some issues with the connection to the All in all, an abstraction of the MAVLink protocol
corresponding software component, right away. More simplifies the usage of this protocol drastically. Using the
specifically, the MAL executed the issued commands as well MAL, a user can easily write a MAVLink command sequence
as provided the requested data in a fast manner. without the need to understand the underlying MAVLink
commands executed or the need to evaluate responses from
TABLE I. IMPLEMENTED ABSTRACT COMMANDS
the FLC on their own. Lower level changes can also be applied
Commands Description more easily. While in the absence of a MAVLink abstraction,
№ such changes have to potentially be applied several times, the
1 Alt Hold Change the flight mode to ALTHOLD MAL only needs to be changed once. For example, if the
2 Guided Change the flight mode to GUIDED command to move the UAV had to be changed so a different
3 Land Change the flight mode to LAND underlying MAVLink message is to be used, only little
4 Pos Hold Change the flight mode to POSHOLD changes need to be applied to the MavlinkAbstraction
5 RTL Change the flight mode to RTL (or specifically to the move request). For the user, the
6 Stabilize Change the flight mode to STABLIZE abstraction interface does not change in any way. That being
7 Arm Start the motors said, the biggest disadvantage of the MAL is the added
8 Disarm Disarm the motors complexity, both with respect to efficiency and maintenance.
9 Kill Force disarm the motors Using the MAVLink protocol directly to issue commands and
10 Take-off Take the UAV from GROUND to AIR request data is, if done correctly, more efficient than using a
11 Beep Emit a sound
complex abstraction interface. This is because the user then
12 Move To Move to a target location (latitude, longitude, altitude
has more control over the request made. Furthermore, the user
then may only evaluate the responses necessary to be
13 Move By Move by an offset (forward, rightward, upward)
evaluated. Additionally, there is no overhead added by the
14 Get Device Data Retrieve the flight data of the UAV
abstraction classes and threads.
15 Pixhawk Reboot Reboot the Pixhawk
While the fifteen implemented abstraction functions were
enough for the simple Expert System provided, in the future
After a MAVLink command has passed the simulation further abstraction functions are desired. For example, a
phase, it then has been tested in an indoor environment. More function to circle around a specified point may be useful.
specifically, a UAV has been placed in a cage approximately Furthermore, the move commands implemented do not
the size of 4x4x3 meters. Most of the time if a MAVLink provide movement speed parameters so the exact speed cannot
command passed the simulation phase, it also passed the be controlled. However, this was mostly because the FLC
indoor test without any problem. In cases where the indoor test currently does not support such arguments. In addition to these
has failed, the parameters of the indoor UAV differed from the functionality aspects, another issue of the implemented
simulated UAV. For example, the indoor UAV had its safety MAVLink abstraction should be addressed in the future,
checks enabled while some of these were turned off for the especially if it is expanded upon. The
simulated UAV. Therefore, arming sometimes failed the MavlinkAbstraction class provides the user, if
indoor test. This however does not represent an issue with the requested, with the flight data of the UAV. At the moment,
MAL and was therefore disregarded. this is done via a single abstraction function. However, this
abstraction function internally requests several messages from
the FLC in sequence. While this is done so the user receives
all flight data at once, this also leads to the issue that all of
these requests must be handled by the FLC regardless of if the
user actually uses the corresponding data. A single flight data
request may approximately take around 40ms to be
completed. In the future, this flight data request should be
separated so the user has more control of which data he
requests and therefore is able to limit the amount of data the
Fig. 5. Test environments. FLC has to process.

As the last step, the MAL has also been tested in an REFERENCES
outdoor environment once. Here, only the Test Vehicle has
been used. Because the abstraction commands were already [1] U. Tudevdagva, B. Battseren, W. Hardt, S. Blokzyl, and M. Lippmann,
validated in the indoor environment, they worked as expected “UAV-based Fully Automated Inspection System for High Voltage
in the outdoor environment, too. It is worth mentioning though Transmission Lines.”
that the used thresholds to check whether a flight command, [2] H. Ariane, H. Reda, B. Battseren, and W. Hardt, “AREIOM: Adaptive
Research Multicopter Platform,” IFOST, pp. 219–223, 2019.
e.g. a take-off or a move command, has been successfully
[3] B. Battseren, U. Tudevdagva, and W. Hardt, “A Finite State Machine
executed had to be adjusted. In the indoor environments, these Based Adaptive Mission Control of Mini Aerial Vehicle,” Embed.
commands were executed precisely, meaning that there was Selforgan. Syst., vol. 5, no. 1, pp. 14–18, 2018, doi:
only little noise in the resulting UAV data. However, because 10.14464/ess51225.
of external influences in the outdoor environment, for [4] Hardkernel co. Ltd., “ODROID-XU4: User Manual.” 2015, Accessed:
example wind, the UAV’s position temporarily had a larger Aug. 05, 2020. [Online]. Available: https://magazine.odroid.com/wp-
distance to the expected position than expected. content/uploads/odroid-xu4-user-manual.pdf.
[5] MAVLink, “MAVLink Developer Guide.” https://mavlink.io/en/
(accessed Aug. 05, 2020).

48
[6] “DroneKit.” https://dronekit.io/ (accessed Aug. 06, 2020). [8] V. Ermakov, “mavros - ROS Wiki.” http://wiki.ros.org/mavros
[7] Python Software Foundation, “Python Language Reference.” 2020, (accessed Aug. 08, 2020).
Accessed: Aug. 08, 2020. [Online]. Available: [9] MAVLink, “C (mavgen) - MAVLink Developer Guide.”
https://docs.python.org/3/reference/. https://mavlink.io/en/mavgen_c/ (accessed Aug. 06, 2020).

49
 Comprehensive Professional Competence Study of
Legal University Professors
Narantuya Lkhundev Sodov Tsedendamba
Department Law of the Orkhon University Development center of Orkhon University.
Ulaanbaatar, Mongolia Ulaanbaatar, Mongolia
[email protected] [email protected]

Hundreds of years ago V.I.Vernadskii noted that the

Abstract- We have clarified further objective and analysis university has a form to organize scientific work and it has
requirements to lecturers who teacher law in the university, and become one way to access to culture of whole humanity
how to include and implement them in official legal acts. It’s a through higher education.
challenge to foster in higher education of any professional area
and level, the right proportion of competency of the subject This presentation is about the findings of a research on
matter, foundational competency and professional competency. the academic activities of law school students and teacher-
We have administered questionnaires among law lecturers of professors, with focus on the foundational competency
National University of Mongolia, the Internal Affairs results.
University, Ikh Zasag University and Orkhon University on Training programs should incorporate a set of
their training, research and professional practice, and made a competencies that take into account the complexity of
chronometrage of their daily work. knowledge, flexibility of methods, critical thinking, market
awareness and current developments, demand based on the
Keywords-competence, law teacher in the university, psychological characteristics of the individual, which will
performance indicators, Objective indicators of pedagogical work form the fundaments for proper decision making ability [1].
The comprehensive competencies of graduates of higher
professional education consist of two parts: firstly, they are
I. INTRODUCTION necessary for professional activities, and secondly, they are
necessary for success in life and activities (development of
One of the main challenges of our education sector, social skills and personal development) [2]. The bachelor's
especially for the training and development of University degree graduate is prepared primarily on the basis of five
professors who train various specialists, is the development of groups of subjects: socio-personal, general academic,
need based training policy for pedagogical activities. Many instructional, professional and professional.
universities in our country are focusing on becoming research
universities from training university at this time. Study and II. THE PAINT PART
modeling of the operations of university professors who train
specialists in various fields and reflecting them in their job The status, activities and workload of lecturers in higher
requirements and qualification definitions has effect on education institutions have been included by the examples of
improving the competitiveness and flexibility of graduates, as trainings, research of law teachers. It has become a priority
well as their professional skills. for higher education to research into the characteristics of the
profession, duties and status of professors at higher
Vocational professors of university master the education institutions by developing the ergonomic basis for
content and develop methodology of teaching, as well as do psychological diagnosis, prognosis, personal development,
scientific research. On top of which he/she has to have improving graduate-professionals capacity, and intensifying
knowledge and complex competence (researching, training, the process using market incentives.
personal) approach to education, psychology and One of the challenges of law education is to train lawyers
professionalism of pedagogy, while consistently continue self with the basic competencies to meet the requirements of a
development. broad and competent application of law norms in the law
The university around the world has quality services market. A law teacher is unique in the sense that
functions including to prepare specialists who work as their he/she is a transmitter of law culture – the regulatory aspect
profession, to build a person, to organize training process, to of social life, acquires the content of educational values based
prepare people who has liquidity in any fields, to organize on professional and moral qualities, transmits professional
training and research among the students and teachers, to knowledge to students in an understandable way, evaluates
prepare specialists in scientific organizations. Therefore phenomena correctly, and actively influences them with law
teachers who work in university have features that to awareness. He/she should be able to diagnose the work of his
research and teach students, masters and doctors or her students, plan the learning process to work
simultaneously. individually, perform it independently, plan activities,
formulate goals, use a variety of advanced learning
technologies to implement them, and have the following
professional competencies [3]:

50
 x Cognitive - to reflect the knowledge, skills and professors and law enforcement experts is very important in
activities of educational institutions in teaching the developing new teaching methods and taking into account the
subject in accordance with the law norms, and to needs of students.
link the training with the profession. The professional description of a law teacher
x Psychological-pedagogical - basic knowledge of includes the psychological norms and professional
psychology and pedagogy of higher education, requirements for his/her professional activities, the essential
ability to work with students individually, ability to nature of the profession shall include the content of the
conduct educational activities based on their activities that define his/her professional goals. However, the
diagnosis model shows the comprehensive professional competence of
x Management - the ability to manage social processes law teacher by reflecting the type of professional activity and
and educational systems, processes, and to analyze qualification of the teacher in the qualification as shown in
and organize the learning process Figure 1 [6].
x Problem solving - proposing and solving problems
in planning professional activities and evaluating
results
x Information - the ability to search, process and use Comprehensive professional
various information in modern information competence of a law teacher
technology tools in professional activities
x Communication - knowledge and skills of written
communication, speaking and pedagogical
Objective Subjective Psychological
communication. indicators of indicators incompatibility
pedagogical
At the same time, self-directed learning, humanization of work
education, democratization, competitiveness, intellectual and performance Quality of
moral freedom to grow as a professional teacher, Professional indicators psychological
independence from others, intellectual strength to overcome target activity
life's challenges, law and educational changes. Nowadays,
students are required to have a high level of intellectual and Knowledge Psychological
cultural awareness, to teach their research skills in the field of Professional quality
duties Skill
law, and to constantly improve their pedagogical,
psychological and pedagogical skills. Professional
We have included a total of 68 teachers from the Law The result Habits estate
of
Enforcement University, the Law School of the National educational
University of Mongolia, the Law School of the University of work Axiological
Mongolia, the Law School of the University of Ulaanbaatar, invariants
the Law School of the Ikh Zasag, and the Law School of the
Mongolian National University; in the research for Fig 1. Comprehensive professional competence of a law teacher
information acquisition, processing, training, vocational
training, and self-analysis. Table 1 shows the self-assessment According to our many years of research, there is a need
of these skills by professors [4]. to ensure the effectiveness of pedagogical activities in tertiary
law education in front of the universities of our country, and
there is no system for objectively evaluating the activities of
TABLE I. ASSESSMENT OF ONE'S ABILITIES BY LAW TEACHERS
professors and teachers; there is also a conflict between the
А. Have you mastered these skills? need for their professional development and the ability to
Question Б. Did you develop these skills during the provide a level of expertise. The results of our research show
law research?
В. What skills do you need to prepare to
that there is a need to improve the teaching methodology by
teach? identifying ways to improve it based on a detailed study and
Ability Г. Note your underdeveloped or mandatory analysis.
skills?
А Б В Г Actual indicators of law professional teacher training
Communicate 40.4% 20.6% 4.4% 2.2% (professional goals, responsibilities, teaching results);
Obtain, process and use subjective indicators (performance indicators, knowledge,
33.8% 22.8% 4.2% 6.6%
information. skills, habits, psychological characteristics, estate); There is a
Training 33.4% 22.4% 8.8% 2.4% psychological assessment that is necessary to achieve these
Professional 35.5% 21.5% 7.5% 2.8%
Examine herself 41.0% 18.4% 4.4% 4.0%
goals. Some researchers [3] teach law teachers
methodological skills (giving students the ability to
Explanation: Number in the table shows average percentage of the answer.
summarize and analyze law phenomena); ethical (related to
This assessment shows that law teachers are well- accountability for professional and social actions);
versed in these skills, and that they have sufficient training, communication skills (stabilizes law relations based on verbal
professional training, and skills to be more mature and useful and non-verbal communication); We are pleased that the
when preparing for law research and teaching. However, the ability to work with sources of information (training in the
ability to obtain and process information in common foreign use and efficient organization of information resources in the
languages using modern information and communication field of law relations) should be provided through additional
tools is underdeveloped or needs to be further developed. training.
The study of how to build the basic competencies of law The Law professor of the University must have at least a
professionals has shown that the cooperation of universities, master’s degree in Law, mu have theoretical knowledge and

51
practice in teaching the subject, mastered research was that the cooperation of universities, professors and law
methodologies, possess knowledge of any common foreign enforcement specialists is very important in developing new
language at a level of using professional books and literature teaching methods [5].
and as well as skills and capabilities to develop methods to In order to eliminate the discrepancy between the need to
teach the materials, must have abilities to use communication ensure the effectiveness of pedagogical activities in
technologies and tools, must be able to follow the ethics of a Mongolian universities, the lack of a realistic system for
lawyer and a professor, and have professional and polite evaluating the activities of professors and teachers, the need
communication skills. In addition, the following knowledge, for their professional development and the possibility of
skills, competencies, and characteristics are required. ensuring the level of professional evaluation, professors of
Including: some Russian universities - Developed a standard for
Knowledge: Science-based foreign language, cyber teaching professors and a standard of bachelor's degree in law
information and communication technology curriculum approved by the Bachelor of Law Standards and Metrology in
design, development of new methods for technology, Resolution No. 36 of 2007, and developed a standard for
pedagogy, general information and culture. university professors and teachers, and the average number of
Skills: Communication, public speaking, creativity, civil servants for senior lecturers. It is recommended that a
collaboration, research, methodological, correct and effective salary commensurate with the salary be based on a credit hour
expression of ideas, independence, patient acceptance of system.
what students say and write, discernment.
Competence: Responsibility, ability to withstand
pressure from critics, independent research, basic ACKNOWLEDGMENT
professional competence of subject research, critical We would like to thank the members of the Symposium
approach to the work of others and oneself, abstract thinking, Organizing Committee and all those who provided in-depth
identification of key issues in disputes, accurate assessment assistance in conducting the research and writing the
of phenomena, an appropriate use of cultural, ethical, and presentation.
scientific research methods and results in teaching.
Characteristics: Emotional self control, observant, REFERENCES
exploratory, humane, principled, interest in the profession, [1] N.V.Korniechenko, “Competitiveness of the future specialist in the
personal discipline, self-confidence, curiosity, sociable, legal labor market conditions”, Higher education today, 2008. vol.10, pp.72-
awareness. 74.
[2] V.A.Devisilov “Competence-based Higher Professional Educational
Education Standards: Structure and Content Issues”, Higher education
III. CONCLUSION today, 2010. vol.9, pp.18-22.
[3] V.D.Shadrikov “New specialist model: innovative training and
The law teachers of the surveyed universities have good competence-based approach”, Higher education today, 2004. vol.8,
communication, information skills, self-examination, pp.26-31.
professional training and teaching skills, and sufficient [4] L.Narantuya, Ts.Sodov, “Implementation, objectives, requirements of
training skills, professional training and skills are better than law teacher”, 2019. vol.08/156, pp.28-37.
when conducting law research and teaching; mature and [5] L.Narantuya, “Some issues of university professor-teacher evaluation”,
useful. However, the ability to obtain and process information Research conference of ОU, 2010. vol.11 pp.165-175.
in common foreign languages using modern information and [6] L.Narantuya, Ts.Sodov “Basic Competence Study of Law Students and
Teachers”, “Challenges of Didactics in Preschool, Primary, Secondary
communication tools is underdeveloped or needs to be further and Higher Education” Proceedings of international scientific
developed. conferences, 2015. pp.144-150.
During the research on how to build the basic
competencies of law professionals, one of our observations

52
 Assessing the Progress of Children’s Competencies

Narantuya Jamyandorj Ganbat Tsend

Department of Science and Engineering School of Information and Communication
University of Mandakh Technology,Mongolian University of Science and Technology
Ulaanbaatar, Mongolia
Ulaanbaatar, Mongolia
[email protected]
[email protected]

Abstract—The purpose of this paper is that it can be used

as a model for assessing the ability to understand and
perform tasks of preschool children and compare them with
other children, determining their progress in basic
education, and estimating how they change with age and
gender. The research was conducted among 60 girls and
boys in a kindergarten. The method developed in this study
suggests that training and play programs for children should
be developed for each age and gender.

Keywords—Child and computer interaction; Child and

cognition; Machine learning

I. INTRODUCTION

Pre-school education in Mongolia aims to develop

young children through their own characteristics, skills,
and creativity and to provide them with basic pre-school Fig. 1. A task to switch a shape on a smart phone.
education. The content of the training includes the
acquisition of language or speaking and thinking skills, the II. RESEARCH DATA COLLECTION
development of drawing and creativity. The basic teaching
tools are visual materials and games. Procedures have been Sixty male and female pre-school children aged 3 to 5
adopted to assess children’s abilities, competencies, and years (10 males and 10 females in each age group) were
developmental progress. randomly selected and instructed by the teacher, the time
Pre-school education covers all children from the age for understanding and performing the task were recorded
of two to the time of enrollment. Kindergarten training is on the smart device [7, 8].
organized in primary, secondary, senior, and preparatory Task: Switch the “red” figure to the “green” one.
groups. Kindergarten is staffed by professional teachers. Before giving the task, the teacher records the child's
The head of the kindergarten has the right to test and age and sex on a smart device. As soon as the teacher has
introduce advanced methods and forms of teaching finished telling the child the task, the teacher records the
activities, and the teacher has the responsibility to discover time (su) when the child begins to think. After the child
and develop the talents, skills, and interests of the children. listens to the task, the device records the time (eu) that
takes to finish thinking as soon as he places his finger on
The core curriculum of preschool education states that
children must learn through all the senses gaining the screen.
experiences. This includes listening to the teacher's If we consider the time to understand as u_time, it will be
instructions with the ears, seeing with the eyes, searching as follows:
in the memory, reflecting and pointing to, holding, and u_time = eu - su (1)
moving the given material. For example, brushing teeth
with a brush, combing hair with a comb, drawing on a Once the child has switched the red shape to the green
paper with a pencil, painting shapes with colored pencils, one, the device will record the time (em) that takes to
picking up toys, etc. The teacher keeps a brief observation complete the task.
note on each child. It is also necessary to evaluate the If the time for performing the task is denoted as m_time, it
progress of each child and compare the overall assessment will be as follows:
of the children in the class [10, 11]. m_time = em-eu (2)
The teacher can give a task a child to draw the red dot
on the paper and connect it with the green dot using a The following table shows the identification numbers of
pencil. In this case, the teacher can record the time the the first 30 children, their age, gender, and the time taken
child understood the teacher's task and the time it took to to understand the task and the time taken to complete the
draw it. Also the children can be given a task to switch a task in (centisecond).
shape from red to green using a smart phone. The authors
conducted the survey data using smart devices (Fig.1.)

53
 TABLE 1. Research data

ID age gender u_time m_time

1. 3.1 male 4089 245 Fig. 2. The similarity of the two vectors will be found at angle.
2. 3.1 male 625 224 (sim(d1,d2)=cosⱷ)
3. 3.3 male 8741 547
4. 3.1 male 5535 252 The similarity of the two vectors in vector space will be
5. 3.1 male 4256 285
6.
found with the formula of cosine similarity.
3.1 female 2272 505
7. 3.3 female 4548 165
8. 3.6 female 327 239
9. 3.6 female 2197 446
10. 3.6 female 1127 247 (4)
11. 4.6 male 1418 293
12. 4.3 male 2853 125
13. 4.4 male 1301 770 Here V(d1) - d1 coordinate of a vector
14. 4.5 male 1229 298 V(d2) - d2 coordinate of a vector
15. 4.4 male 5048 190 |V(d1)| - d1 length of a vector
16. 4.5 female 4699 497
17. 4.5 female 1379 463
Using these two formulas shows how different the time of
18. 4.3 female 1485 311
19. 4.10 female 1375 626 understanding and performing the task depending on
20. 4.3 female 325 347 children’s age and gender.
21. 5.6 male 1348 201
22. 5.3 male 1003 330
23. 5.6 male 3492 670 IV. PROCESSING
24. 5.7 male 5389 368
25. 5.7 male 1580 275
26. 5.3 female 4585 209 The round blue shapes are 3-year-old boys and the
27. 5.8 female 856 277 triangular green shapes are 3-year-old girls; the
28. 5.5 female 1759 139 rectangular shape is the centroid.
29. 5.4 female 1766 778
30. 5.3 female 2155 284

III. METHODOLOGY

A vector space can be created on the two axes of time

u_time and m_time to represent the time that each child
will understand and perform the task in vector space. Each
child can be considered as a vector with two coordinates.
Rocchio’s formula was used to find the average time to
understand and complete the task for each age and gender.
It was developed by Rocchio in 1971 and is a formula for
finding the centroid coordinates of points in vector space Fig. 3. Boys aged 3, their centroid.
[2].

(3)

V(d) - Coordinate of d vectors

|Dc| - the length of the corresponding coordinates of the
points

The cosine similarity formula for finding the similarity of

two vectors was used to find out if the competencies of
the two children were the same within the class.

Fig. 4. Girls aged 3, their centroid.

In the same way the average centroid of boys and girls

aged 4 and 5 was found.
With the centroid, children can be divided into the
following four categories by their age and gender.

54
 1. Children with good comprehension and performance
2. Children with good comprehension, but slow
performance
3. Children with slow comprehension, but good
performance
4. Children with both slow comprehension and
performance
If we mark as cx and cy according to the coordinates of
the centroid, the above 4 categories are defined as follows:

1. Class C1 of points with (x, y) coordinates that meet

the condition of 0<x<cx and 0<y<cy. Here will be
involved the children who can understand and
perform the task successfully.
2. Class C2 of points with (x, y) coordinates that meet Fig. 6. Boys aged 3-5, their centroid.
the conditions cx<x and 0<y<cy. Here will be
involved the children who understand the task In the same way, the progress of boys' competencies
slower, but perform it successfully. was determined with a graph. The following table shows
3. Class C3 of points with (x, y) coordinates that meet the age and gender differences for children who
the conditions 0<x<cx and cy<y. Here will be successfully understood and completed the task.
involved the children who understand the task
successfully, but perform it slower. TABLE 2. Age and gender difference and changes
4. Class C4 of points with coordinates (x, y) that meet
the conditions cx<x and cy<y. Here will be involved Age, gender P(B|A)
1. All children aged 3 0.4
the children who understand and perform the task 2. All children aged 4 0.5
slower. 3. All children aged 5 0.6
4. Boys of all ages 0.59
5. Girls of all ages 0.48
6. Children of all ages 0.63

Judging from this, children’s competencies are different

with their age and gender.
VI. CONCLUSION
The authors have come to the following conclusion after
this research work.
1. For 3-year-olds, girls have a higher ability to
understand and perform tasks than boys, but with age,
boys are able to understand and perform faster than
girls.
Fig. 5. Boys and girls aged 3, their centroid.
2. Boys are slower to understand a task, but they
perform better than girls.
With the data of the 20 children aged 3, the probability of 3. This study is not an assessment of a child, but a
entering the four categories divided by h children was measurement that is taken to determine a child's
calculated. cognition level.
Except for using centroid, there can also be used
P(h|C1)=8/20=0.4 achievement scores gained by all children to determine
P(h|C2)=4/20=0.2 children’s competencies.
P(h|C3)=2/20=0.1
P(h|C4)=6/20=0.3
REFERENCES
In addition to this, the probability for the children of other [1] I.Scott MacKenzie, Fitts Law as a Research and “Design Tool in
Human – Computer Interaction”, 1992.
remaining ages was calculated. As children got older, their
[2] Rocchio, J.J.: Relevance Feedback in Information Retrieval. The
chances of getting into C1 increased and their chances of SMART Retrieval System, pp.313-323, (1971).
getting into C4 dropped. [3] Elomaa, T. & Rousu, J. (1999). General and Efficient Multisplitting
of Numerical Attributes. Machine Learning 36, pp. 201–244.
[4] Auer P. & Warmuth M. (1998). Tracking the Best Disjunction.
V. RESULTS Machine Learning 32, pp.127–150.
The graph shows the progress of boys' competencies. [5] Breiman L., Friedman J.H., Olshen R.A., Stone C.J. (1984)
Classification and Regression Trees, Wadsforth International
Group.

55
[6] S. B. Kotsiantis, Supervised Machine Learning: A Review of [9] Ts. Ganbat, D, Sarangerel, “Human and Computer Interaction”,
Classification Techniques, pp. 249-268, 2007. Ulaanbaatar, Mongolia, 2018,
[7] Аnry Vallon, “Formation of characters”, 1934. [10] Law on preschool education, Laws of Mongolia. 2008.
[8] J. Batdelger, “A test to determine the general level of mental [11] Order # A/327, Policy on preschool education, 2013 .
development of a child”, Ulaanbaatar, 1996в

56
 Analysis of User Interface design methods
Nikolai Gervas
Computer Engineering department
Novosibirsk State Technical University
Novosibirsk, Russia
[email protected]

Abstract — The article tells about the main approaches and

methodologies for UI design, as well as the design process in
general. Basic methodologies usually describe UI design in terms
of an ergonomic aspect, rarely consider the features of
functional design. Therefore, the article deals with the analysis
of requirements and the role of the functional aspect in UI
design. The decision to research this topic arose while
developing its own web-application “Fireproof Corporation”.
Therefore, the considered features of functional design will be
accompanied by own examples from this application.

Keywords — User Interface, analysis of requirements, design,

aspect, scenario.

I. INTRODUCTION
Actual problem for any web-application is the design of an
effective interface, where efficiency is understood as
simplicity of development, ease of maintenance and working
with the program. There are enough standard methodologies Fig. 1. Scheme of GOMS methodology.
and recommendations for UI design. However, most of the
research is on the ergonomic aspect. The important point in UI Due to the fact that Machine-Centered approach to
design is the functional aspect. The user interface design interface design is focused on the functional characteristics of
process and its relationship to systems analysis (requirements the program, the user working with it is forced to “think like a
analysis) should be investigated. The research began with the developer”.
development of its own web-application “Fireproof Human-Centered approach, which replaced algorithmic
Corporation”. During the development, several design modeling, considers the user as the central figure in the
oversights were made, which were corrected and analyzed. In process of interacting with the system. Orientation to the
the article, it will also be considered from the point of view of characteristics of the user, the study of perceptual and
the functional aspect. cognitive capabilities and limitations of a person made it
II. ANALYSIS OF UI DESIGN APPROACHES AND possible to identify patterns of human interaction with an
METHODOLOGIES
automated system. Considering the processes and patterns of
perception, information processing and decision-making,
First of all, when analyzing UI design methodologies, cognitive psychology identified the factors that determine the
design approaches should be considered. There are two success of the operator's task. And these turned out to be not
approaches to UI design: Machine-Centered and Human- the functional characteristics of the system, as it was assumed
Centered [1]. These two approaches, in fact, represent an by engineers earlier, but the quality of the provision and
automated system at the highest level of detail ("human || management of information in terms of human capabilities
computer") and consider the interface development process and limitations.
either from the standpoint of a human operator or from the
side of the functionality of a computer. However, as it turned out, the analysis of only the
processes of perception and processing of information by a
Machine-Centered approach to creating a UI is based on person is not enough to design an ergonomic interface, since
the assumption that a person works with a computer like the it does not allow determining the composition and sequence
computer itself, i.e., according to a certain algorithm. of information displayed on the screen. This has led to the
The algorithmic modeling technique GOMS ("Goals - emergence of a number of cognitive-based UI design
Operators - Methods – Selection rules"), representing this methodologies.
approach, assumes that the result obtained when the user Activity-Centered Design. This methodology considers
performs a certain task is a goal. To achieve it, the user can the "man-computer" system as a complex of related activity
perform elementary actions - operators. A sequence of concepts and representations. The theory of activity
statements that allows you to achieve a goal is called a underlying this approach presents the computer as a tool with
method. Selection rules based on the "if-then" principle allow which a person solves various problems and it is human
changing the control flow. The structural scheme of this activity that affects the interface [1, 2].
technique is shown in Fig. 1.
According to the principles of activity theory, the entire
stream of user activity can be decomposed into a sequence of
related tasks and subtasks, logical stages. This allows to
analyze goals, external and internal tasks, the order and type
of user operations performed to achieve the final result, and

57
based on the results of the analysis, develop an interface that forms of the interface and the sequence of manipulations with
is most suitable for this type of activity them. It is difficult to talk about the quality of the interface
functionality regardless of the subject area, for example, to
Goal-oriented design. This user interface design formulate the "guidelines for the functionality" of the user
methodology, which Alan Cooper is the ideologist of, is based interface. Formally, it can be associated with the degree of “fit
on the premise that a careful study of the user's goals and for purpose” [4]. This is described in the ISO 9241-10-1996
understanding of what they are aiming for can solve the Ergonomic requirements for office work with visual display
problem of "cognitive friction" [1, 3]. Cognitive friction is a terminals (VDTs), p. 10, Dialogue principles. Referring to this
concept introduced by A. Cooper and characterizes a person's same standard, p. 11, Guidance on usability specification and
attitude to a complex thing (for example, to a computer) as to measures, we can assess the quality of the UI using three broad
another person. This attitude arises in situations where a criteria:
person cannot understand how and why this thing works (or
does not work). x effectiveness - the effect of the interface on the
completeness and accuracy of the user's achievement
User-Centered Design. The methodology has gained
of the target results.
justified popularity and is applied not only in software
development. Its essence boils down to studying the needs and x efficiency or the effect of the interface on user
capabilities of end users and adapting the product to their productivity.
needs. In other words, it is the concept of creating products,
including software, that people would like to use. x the degree of subjective satisfaction of the user with
this interface.
All the considered methodologies have a design process
similar to each other. User interface design usually consists of Efficiency is the criterion for the functionality of the
several main consecutive steps [7-8] shown in Fig. 2: interface, and the degree of satisfaction and, indirectly,
productivity is the criterion for ergonomics.
When talking about the functional aspect of UI design, we
need to consider such an important detail as requirements
analysis. As described in the ISO 13407:1999 (Human-
centered design processes for interactive systems) standard,
user-centered design begins with a deep understanding of the
needs and requirements of users. Methods for collecting
requirements are not considered in this article, as it has been
Fig. 2. UI design process structure. extensively researched. The main interest is processing
requirements and functional interface design based on it.
x Research - collecting data about users of the product.
Scenarios. All functional requirements are presented as
x Modeling - behavioral patterns identified through scenarios. A scenario is a description of actions performed by
analysis are collected in the form of domain models. a user in the framework of solving a specific task on the way
x Development of requirements - analysis of data to achieving his goal. Different types of scenarios are used at
associated with characters, their functional needs and different stages of goal-oriented design, and at each
interaction with other characters in various contexts. subsequent stage more attention is paid to interface features
than at the previous one. The first type of scenarios are
x Definition of infrastructure - the creation of a common contextual scenarios created during the initial design stage,
product concept, a concept of behavior, graphic design written from the point of view of the character, and focusing
and physical form. on human actions, impressions, and desires. Next comes the
development of key path scenarios that focus on the most
x Detailing - similar to the infrastructure stage, but more important moments of the interaction, without losing sight of
focused on implementation details. Completion of how the character uses the product to achieve his goals.
design work. Throughout the process, the design team uses test scenarios to
x Development support - adjustment of design decisions. test design solutions in various situations. Based on the
identified scenarios, the structure of screens is developed, the
For the purposes of this article, we are interested in the number of screens, the functionality of each of them,
requirements analysis process and its relationship to UI navigation links between them is determined, the structure of
design. the menu and other navigation elements is formed [6].
III. ANALYSIS OF REQUIREMENTS. FUNCTIONAL ASPECT For clarity, it is necessary to consider an example scenario.
It has already been discussed in the article that in the The article will look here and further on examples from my
methodologies for designing the UI, ergonomic standards are own developed web-application “Fireproof Corporation”. It
mainly described. In software design methodologies (e.g. should be said that the system is quite complex and contains
Unified Process), UI development is a secondary activity. It is many business entities related to each other. Therefore,
believed that if the functionality is sufficiently developed in scenarios usually involve the use of several core entities. Let's
the form of requirements and scenarios, then UI problems are consider a simple scenario for adding basic information for the
limited by the features of graphic design, ergonomics, etc. [5, “object” business-entity. Fig. 3 shows the main panel of this
6]. This is not entirely correct, because the functional aspect business-entity, which contains basic information about the
also plays an important role in UI design. The functionality of entity.
a software system should be considered the semantics of its
UI. In this case, the role of syntax is played by expressive

58
 positioning. The interface of the “contracts” page in the web-
application is shown in Fig. 5.

Fig. 5. Interface of the “Contracts” page.

Fig. 3. Scenario for changing counterparty and address.

The administrator manually fills in all the fields with The design of the general structure consists of two parallel
information: name, address, counterparty, etc., which is a processes: the allocation of independent blocks and the
painstaking process. It should be noted that both business- determination of the relationship between them. The
entities "object" and " counterparty" have a specific address. allocation of independent blocks is performed based on the list
We know from the system requirements that the addresses are of individual functions identified using user scenarios. In the
most often the same. Thus, we can get rid of an unnecessary application, a function is represented by a function block with
action - the constant filling of the object address. It is enough a corresponding screen form. Typically, several functions are
to implement the interface in such a way that when the combined into one function block. On the ergonomic side,
appropriate counterparty is selected, the address will be set functional blocks on the page should be visually separated
automatically. With some exceptions, when the addresses of from each other. This is done using colored rectangular
counterparties and objects do not match, the address can be blocks.
manually changed. Fig. 3 shows that when the appropriate Navigation. The navigation process should be designed to
counterparty (1) is selected, the address (2) is automatically help users determine where they are, where they have been,
set for the object. With the help of such scenarios and and where they can go in the future. It is difficult to give any
requirements analysis, we can get rid of unnecessary steps in specific recommendations for this work, since a lot depends
the early stages of UI design. on the designed system.
General structure design. Once we have defined the
scenarios, we move on to designing the overall structure of the
system. To do this, it is necessary to highlight individual
functional blocks and determine how exactly these blocks are
connected to each other. A separate functional block will
mean a function or a group of functions related by purpose or
scope.
Fig. 4 shows typical structures of common sites and
complex technical systems. While sites are usually forked, in
the sense that functions are usually placed on separate screens,
then complex applications usually have only one changing
screen, in which almost all functions are called.

Fig. 4. Typical site structure (left) and complex application (right).

The implemented web-application “Fireproof Fig. 6. Positioning from the “object” page to the selected application.
Corporation” has the structure of a complex application shown
in Fig. 4 on the right. This UI model was named “The
cockpit”, because a large number of functionalities is located In the developed application, navigation is implemented in
as compactly as possible on one page. Also, the advantage of the simplest way in the site header. Each page (large
this interface is the use of context-sensitive elements and functional block) has a corresponding tab. Based on the fact

59
that the system interacts with a large number of related requirements analysis that the created applications are usually
business entities, a mechanism for positioning the business distributed immediately. Therefore, when saving the
entity of interest to us is implemented (shown in Fig. 6). application (2), we automatically position ourselves on the
Suppose we are on the “object” business-entity page (1) and page of the calendar planner. We also keep the context and
want to view one of its “application”. Choosing one among all the applications the one that we just created is
application from the list, we automatically go to the page of selected (3). Everything is automatically ready for
the “application” business-entity (2) and can view the already distribution.
selected information about this application.
СONCLUSION
Without the implementation of such a positioning
mechanism, the user would have to independently go to the As a result of the work, the main approaches and
“application” page and look for the application of interest in methodologies used in UI design were considered. The
the list. methodologies deal mainly with the ergonomic aspect,
although the analysis of requirements and user needs is also
Chains of scenarios to execute. In a complex technical taken into account in all the methodologies discussed. Despite
system, positioning between pages can be performed within a this, not enough attention is paid to the functional aspect in UI
complex scenario. Requirements analysis identifies which design. In the software design methodology, UI development
scenarios are typically executed sequentially. When you is a secondary task, it is considered that it is enough to work
combine these scenarios, you can understand that some steps out the functionality. The article reviewed the process of
are unnecessary and will complicate the use of the program. requirements analysis, scenario development, the role of the
Consider a typical scenario: we have an object in the functional aspect in UI design. Using examples from the own
maintenance where repair work is to be carried out. Therefore, developed web-application “Fireproof Corporation”, the
we have the following sequence of actions: features of functional design identified during the analysis of
requirements were considered. It is difficult to talk about some
x creating an object in the system. fundamental rules when designing the functional part of the
x creating an application for this object. UI, it is necessary to start from a specific subject area. Despite
this, the developed set of rules and recommendations can help
x distributing an application using a calendar-scheduler. designers avoid obvious mistakes.
It is necessary to simplify this process with automatic The next step towards developing own methodology is to
positioning. A visual representation of the chain of action is consider the relationship of UI design with the main
depicted in Fig. 7 as the corresponding business entity pages. architectural patterns of the MVC / MVP / MVVM.
Information can be found about the functional aspect in
various methodologies, but the consideration of the
relationship between architectural patterns and specific types
of UI is not given in the methodologies, therefore this is a new
research and approach. Certain types of UI should not be
implemented using certain patterns, this will lead to system
complexity and possible errors. This study contains an
engineering approach and is aimed at helping software
developers directly.
REFERENCES
[1] A. Anatolyev, “Design of human-machine interfaces,” 2014. [online].
Available: http://www.4stud.info/user-interfaces/
[2] D. Holmes, “Activity Centred Design,” 2018. [online]. Available:
https://medium.com/dermot-holmes/activity-centred-design-
dd28ed1eec59
[3] Qubstudio, Design Agency, “How to Design Products with Goal-
Centered Design,” 2018. [online]. Available:
https://qubstudio.com/blog/how-to-design-useful-products-with-goal-
centered-design/
Fig. 7. Scenario for creating an application for an object. [4] E. Volchenkov, “User interface standardization,” 2002. [online].
Available: https://www.osp.ru/os/2002/04/181312
First of all, we fill in the newly created object with [5] E.L. Romanov, G.V. Troshina, A.A. Yakimenko, “Teaching Software
information and click on the "Add application" button (1). Engineering. A glance from the code,” vol. 8, no. 5, 2016. [online].
After that, we automatically position ourselves on the page Available: http://naukovedenie.ru/PDF/66TVN516.pdf
with applications, where we have an application that is already [6] Brusentsova, T. P, Designing user interfaces: a guide for students of
automatically created and filled with information (about the specialty 1-47 01 02 "Design of electronic and web publications".
Minsk: BSTU, pp. 172, 2019.
object, time of creation, etc.). Such functional design greatly
facilitates the user's work, since he would have to do all the [7] V.V. Golovach, User Interface Design. Russia, pp. 762, 2000.
automatic work manually. We also know from the [8] S. Krug, Do not make me think. Web Usability and Common Sense.
Moscow, Russia: Eksmo, pp. 229, 2014.

60
 Developing an Information Culture for Teachers
Nyamsuren Аmsarvaa Tsedendamba Sodov
Science section Teacher development center
Orkhon Khasu secondary school Orkhon university
Ulaanbaatar, Mongolia Ulaanbaatar, Mongolia
[email protected] [email protected]

Abstract— Teachers' information culture is becoming II. THE MAIN PART

more important in mastering modern information and
communication technologies and introducing them into The Teacher’s Information Culture is an integral part
training and developing every student. of a teacher's general culture that incorporates knowledge,
Due to the fundamental changes in the teaching methods and skills, and worldviews about information in a global
attitudes of teachers, developing an information culture for
professionals has become an important goal of education policy
networked information society in a professionally oriented,
to build professional knowledge, skills and abilities, as well as health-oriented manner. The teacher information culture
the ability to create and use information resources. includes elements of worldview, cognition, action, and
ecology, and by mastering such a culture, one can
Keywords— information culture, teacher information communicate information in a way that is understandable to
culture, information and communication technology others by working with, structuring, systematizing, and
summarizing it.
I. INTRODUCTION
In general, a teacher's knowledge of computer involves
In today's world where teachers' information culture is computer knowledge (software usage, programming,
becoming more and more important for mastering and educational process planner), ability to use new ICT tools to
introducing modern information and communication solve educational problems, ability to cooperate with
technologies, the development of teachers' information students in developing themselves using
methods and attitudes is fundamentally changing, so telecommunications during the learning process, quick
developing information culture for professionals has become testing of others knowledge, reflecting necessary
an important goal for creating knowledge and skills. information into training. As teachers become more
information literate, their teaching methods improve, as do
An important condition for the transition from an their professional qualities, personality traits, aspirations,
industrial society to an information society is the attitudes, and personal qualities.
development of an important component of open education
for teachers - the information culture. There is a conflict For example, procedures for collecting, processing,
between the need to use information and communication storing, retrieving, using, and disseminating information are
technologies (ICT) in education and the training of teachers specific activities with information objects, so that ICTs not
to use them in the lessons they teach. First of all, this is only enable a person to use information, but also work with
primarily due to the lack of information culture for teachers; information, acquire a moral and ethical culture, and
Second of all, many teachers are not sufficiently prepared to develop cognitive skills.
use new information technology in their teaching; Third of
all, teachers and students are less motivated to use First of all, the ability to use information and
information technology in education. At a time when the information technology; Secondly, the ability to
pedagogical culture of teachers includes a comprehensive communicate with ICTs by selecting what you need from
information-pedagogical capacity, the information culture is the information resources, recognizing the importance of
an orderly system of knowledge, skills and practices of using information in print and electronic form and
ethical norms and cultural-spiritual values that ensure the understanding the conditions that affect the source of the
activities of the person whose object is information [1]. information; Thirdly, an information culture can only be
achieved through a comprehensive set of knowledge, skills,
Although information and information technology and practices that support ICT and human action on moral
courses are taught in all levels of schools in our country, in norms, cultural, and spiritual values. Therefore, the level of
addition to teaching materials, students and graduates need teacher’s information culture is the knowledge of
to be able to use ICT freely in solving many educational, information, information processes, models and
household and professional tasks in today's society. This technologies; the ability to develop and analyze information
time, I wanted to study the teacher's information culture and tools in a variety of activities and to use open information
the current situation in which it is used, and to share with systems (the Internet, its services, and resources) in teaching
you some of my experiences and opinions. is determined by practice. Create an information-intellectual
environment by using the opportunity to create a teacher-site
by posting the necessary information to your students on
your Web-site; there is a need to constantly assess the
comprehensive competencies of students in the field of
61
information culture using computers, local and global psychologists has increased, taking into account the
networks. specifics and principles of information in pedagogy and
psychology. For example, the use of the Internet, e-mail,
The content of a teacher's information culture is related computer tests, spreadsheets, databases to save time,
to the holistic ability to master it. It is interesting and increase the accuracy of calculations, detailed analysis,
interesting to distinguish between ICT-based professional psychiatric diagnosis, editing information using the above
development, pedagogical activities, student education in technology and computer games, multimedia technology,
the context of informatization of education, essential graphic editors, Internet, It is possible to conduct social
professional relations, and cooperation between stakeholders dispatching activities using e-mail, electronic presentations,
in education [6]. According to the results of the survey, we multimedia technology, printing systems, graphic editors,
evaluated the information culture assessment of secondary consulting and enlightenment, and inviting highly qualified
school IT teachers. The ability of the surveyed teachers to specialists to improve the quality of co-created materials
work with information and textual information was 41% on without spending time.
average, and the ability to analyze the information found
was 38%; Ability to translate information in common We studied the pedagogical ability to develop a
foreign languages independently and by computer, to search methodology for using ICT in the learning process using a
the Internet for search engines and specialized sites 34.7%, model developed by I.N.Semenova and A.V.Slepukhin [3]
ability to work with WWW (Netscape Navigator, MS (see Figure 1).
Internet Explorer, etc.) 37.5% %, ability to work with
electronic textbooks, trainers and experts 23.8%, ability to The teacher begins by planning a pedagogical approach
search the library for necessary information from general to the student’s activities based on a study of their students'
and electronic catalogs and specialized journals 26.2%. interests and learning abilities. Practice has shown that it is
necessary to take into account the potential impact of the
The methodology for using ICT in training is: first, teacher, as well as the possible pedagogical goals, possible
ICT in certain circumstances, and the rules for selecting pedagogical assessments and possible conditions for the
these components; Second, the planning, editing, selection, learning process. Experiments have shown that our high
and adaptation of ICTs to specific subjects in a specific school students are very willing to combine laboratory
psychological-educational context; Third is the science of lessons with virtual and real-world experience [5].
applying technology to achieve the learning objectives in a
specific context of psychological-pedagogical content of a We use the activities of teachers to implement ICT and
subject. Therefore, the training to use this ICT is: diagnostic methods:

a. Explain the learning process using ICT a) Teacher knowledge of the pedagogical opportunities
and features of ICT implementation;
b. Analyze existing ICTs
b) The teacher's ability to determine the course of action;
c. Introduce ICT in certain pedagogical and
psychological contexts c) The methodology for using ICT in professional
practice is considered at the following three levels.
d. By creating and modeling ICTs on their own, it is
possible to enrich and develop the methodology for their use After learning to do such activities, a teacher can become
(to provide a pedagogical basis for studying the realities of an ICT trainer, caregiver, and consultant. In addition, the
pedagogy). teacher becomes motivated and creative in all of these
activities. In other words, a teacher with an information
Taking this into account, teachers need to consider the culture logically connects all the stages of modeling,
relationship between content, purpose, principles, methods, planning, implementing, evaluating, diagnosing, forecasting,
activities of educational subjects, and diagnosing and correcting, and motivating learning activities in a logical
monitoring the learning process when developing a sequence to the specific characteristics of their students, the
methodology for using ICT in the classroom. According to educational environment, and the current state of ICT.
I.M.Bogdanovskaya and others [2], the role of ICT in Students will be able to plan and implement lessons
solving professional goals in the activities of educators- according to their abilities.

62
Fig. 1. A functional model of pedagogy to develop a methodology for using ICT in the learning process

63
 conduct activities, to manage their joint activities, to
One of the key factors in informing the learning process organize administrative work), motivator (to encourage
is the teacher's level of information culture, which students to be creative and to acquire knowledge
determines the following unique and important indicators of independently), planner (to plan and organize the learning
the profession: process in the most effective way), can be a facilitator
(facilitator of learning).
x Interest in improving the educational process using
information technology and the need to constantly 2. Create and use an open educational media system
update their knowledge in the field (educational site, Moodle system, ...) related to the psycho-
physiological characteristics of the student-subject studying
x Cautious and humane approach to news and online, living environment, employer's order, and national
information objects on the Internet, critical analysis development prospects. The challenge is to communicate
of information use results and share experiences.

x Ethical, realistic and patient approach to computer- 3. Always keep in mind that students have the right to
network communication develop their curriculum. In addition to the exercises and
assignments given online, the content of the lectures should
x Democratic, ethical and ethical aspects of be changed to suit their interests and needs, and the views of
pedagogical communication and interaction with parents, the media, and other stakeholders should be
others in a virtual environment explored. The challenge for the education sector is to
develop ICT methodologies and technologies in the
curricula of teacher training colleges and in-service teacher
x They will be able to adapt to the information society
training programs, to organize a wide-ranging discussion of
and have professional energy.
teachers' results and to implement them as a matter of
urgency.
Due to the Covid-19 pandemic in 2020, all countries
around the world are conducting online training and taking
IV. REFERENCES
exams at all levels. In this regard, we are conducting some
research and experiments to develop a methodology for [1] T. P. Zaichenko “Information culture as a psychological and
preparing computer presentations and video lectures for pedagogical factor in the development of universal educational
teachers in the field of information culture. activities of students”, 2015.
[2] I. M. Bogdanovskaya, T. P. Zaichenko, Yu. L. Project “Information
technologies in pedagogy and psychology”, Saint Petersburg, 2015.
III. CONCLUSION
[3] I. N. Semenova, A. V. Slepukhin “Methods of using information and
communication technologies in the educational process” Part 2,
1. A teacher working with students can only play the Yekaterinburg, 2013.
role of disseminator of information culture only if he or she [4] M. Delgerjav, A. Adiyasuren, B. Bolormaa “Teacher training problem
constantly develops himself or herself [4]. By cultivating all (survey review)”, Ulaanbaatar, 2020.
the qualities of a teacher, we become a teacher (a source of [5] Ts. Sodov, B. Purevjal “Research on the use of information
technology tools in physics teaching”, Educational research, №5,
professional knowledge, skills, practices, and experience), a Ulaanbaatar, 2020.
counselor (a student who completes assignments online, a [6] L. N. Denisova, L. N. Chernova “Information culture of a teacher and
counselor in accessing Internet resources, and answers renewal of professional education”, Successes of modern natural
questions), and a facilitator in the learning process (to science, No.5, 2011, p133-136.

64
 “Mathshop” E-Learning Mathematics Program -
Experiments And Opportunities For Future
Development
Otgontugs Bayarrintsen Erdenebat Dashtseren Lhagvasuren Purev
Foundation for Education Loan, School of Mathematics and Natural Sciences, Mongolian Institute of Educational Research
University of Finanace and Economics Mongolian State University of Education Ulaanbaatar, Mongolia
Ulaanbaatar, Mongolia Ulaanbaatar, Mongolia [email protected]
[email protected] [email protected]

Bilguunbaatar Tumurkhuyag
Baganuur joint stock company,
University of Finanace and Economics
Ulaanbaatar, Mongolia
[email protected]

Abstract—The “Mathshop” e-learning math program is

experimentally used by 263 randomly selected 12th grade students B. Consumers’ perspectives Maintaining the Integrity of
of high schools. The results of the experiments show that practice the Specifications
with Mathshop could help the participants to achieve better x School and teacher: Conduct tailored training based
performance in the General admission test. on each child’s uniqueness, current level and
Keywords: Covid-19 learning, e-training, MathShop, self- learning speed; Promote confidence and creativity
learning, in children; Promote self-learning skills; teach them
how to use extra training materials; increase
parents’ involvement in children’s learning process.
I. INTRODUCTION
Among many sectors, deeply affected by the COVID-19 x Parents: Support children’s learning processes
pandemic situation, the education sector emerged as one of (signing them up for extracurricular activities and
the hardest hit and therefore required necessary adjustments. helping on the homework; Educate children in a
As per Mongolia, the education sector took steps into an safe and healthy environment; Check on children’s
unchartered, unprepared and unfamiliar territory of “e- grades and performances, regularly.
learning” as replacement of “classroom learning” system, x Student: Learn successfully and stay on top of their
amid COVID-19 situation, starting 2nd week of 3rd semester program; constantly improve their grades; spend
of 2019-2020 school year. leisure time efficiently; always challenge yourselves
It was major challenge for every secondary schools, by comparing with others; pick a subject to study in
universities and colleges. Yet, in order to overcome this an advanced level; compete on a tournament and
challenge successfully within short period of time, tryouts; learn foreign language(s); get admitted to
collaborating with privately-owned training centers that offer an international institution; to fulfill Mongolian
e-learning program was wise and time-saving decision. program requirement if living and studying in a
Herewith, we will discuss how such collaboration between foreign country with parents.
MATHSHOP e-learning initiative and Education department
of Baganuur district benefited their target school children. C. Program developers’ perspectives:
Improve quality of and access to mathematics education
II. DEFINING THE NEEDS OF THE EXPERIMENT for all within the framework of social responsibility; develop
AND EVALUATION respective programs further based on consumers’ desire and
needs AND provide continuous support.
A. Social perspectives OBJECTIVES OF THE EXPERIMENT AND
Following needs emerged due to emerging pandemic EVALUATION
situation namely: try to preserve learning environment and
process as much as could; not to lose behind global a. To introduce modern and innovative learning tool to
education system’s performance; further develop Mongolian Mongolian children and youth
children, intellectually; introduce innovative ways and b. To bolster advantages and benefits of
technologies to educate children; maintain sustainability and “MATHSHOP” e-learning mathematics program during
inheritance of training curriculum; provide equal learning global pandemic situation and quarantine
opportunities; improve access to education; improve
outcomes and quality of education system; help children and c. To support school children who are not familiar
adolescences spending their leisure time, effectively. with the e-learning transformation
d. To study opportunities to further develop and improve
the program

65
 PROGRAM PROFILE that they may have with the program developers.
Such two way communications were instrumental
Meant for school children through Kindergarten to 12th to improve the program.
grades. 0-7 grades have beginner, intermediate and advanced
levels.
IV. RESULTS
а. To familiarize with the basic concepts
Ninety percent of the participants have responded that the
b. To connect with previous knowledge and solve “MATHSHOP” is the new tool and program that enabled
complex problem individual learning techniques, skills and experiences,
provided new knowledge while saving time. It was definitely
c. To solve critical thinking and complex problem by
interesting for students to solve math problem by using
using basic concepts
computer program, even more so, finding out whether they
9-12 grades have only 1 standard level for each. A student is solved it correctly or not at the moment was a rewarding and
considered to be meeting the criteria of each level when is motivating. Students were able to work on more problems
able to solve all problems and tasks. Grade 12 program within less days proved that it was a worthwhile. 14,023 is
features various problems covering 75 different topics and the total number of math problems solved by the participants
most of them can be randomly assigned which makes it during the experiment.
unrepeatable. It also has a section on General admission test.
This interactive program features multiple functions such as A. Total of 263 students’ math skills were tested in
fill-in-the-blank, choose, paint and draw, when solving Baganuur district and their average scaled score1 was
problems. 524.99 (initial score2 32.14). This ranked them in the
1st place, nationwide.
III. METHODOLOGY OF THE EXPERIMENT AND Student B.Batsugar, one of the experiment participants
EVALUATION from “Education” complex school, got the highest score
Total of 10 interviews were conducted among students while student B.Altan-Ochir, out of “Gun Galuutai”
during the experimental period. Such qualitative data complex school took the 2nd place. The Fig. 1 shows the
collection processes were supported by essential steps such comparison of average scores. The average scaled score
as taking notes, making analyses on weekly statistical results among all participants was 604.71 (79.72 score higher than
of solving problems and overall performances of the the district average which was already ranked on top of the
General admission test. nation). The Fig. 2 shows comparison of initial score scores.
The initial score of participant group was 45.33 (13.19 score
A. Collaborating partners higher than the district average). There were 3 students who
have scored over 700, 8 students with over 600, 12 students
Education Department of Baganuur district of with over 500, only 1 student with below 500 scores (484
Ulaanbaatar city, “GUN GALUUTAI” complex school, scores), respectively. The Fig. 3 and Table I show the
“Education” complex school, “OYUNII ERIN” complex comparative results from the last two years.
school, 24 randomly selected students from “Sod” secondary
school (who have expressed their interest to participate and
had prepared, accordingly), their parents/guardians, masters
students from Non-profit organization’s management
program of Finance and Economics University,
“MATHSHOP” Program developers and Senior executives
of MOSS LLC.

B. Duration: 2020.4.1-2020.6.14

C. Phases of the experiment

x Introduction phase: Activities such as identifying
key partners and target students, making them
familiar with the program, get registered and
briefing on users guidelines, took place in this
phase. Fig. 1. Comparison of national average, Baganuur district average and
experiment participants’ average scaled scores
x Monitoring use of the program and provision of
additional support phase: Program uses of randomly
“It was a difficult time for us due to COVID-19
selected students are regularly monitored in this
quarantine, not knowing how to prepare for the General
phase. Number of solved problems, success rate,
and actual spent time were also being recorded, admission test. Fortunately, MathShop program provided us
weekly basis. Students were informed with their detailed tools and methodology of solving various math
own performance scores and follow-up supports problems step by step, in both online and offline settings.
and praises were given, when necessary. Offering bilingual (Mon and Eng) choices were super
helpful for us who were also studying English.
x The phase to introduce users’ comments and
suggestions to the program developers. 1
Scaled score is the score of an individual compared with all those tested.
Participating students were encouraged to document 2
Initial score is the result of solving particular subject theme and task. It is
and share whichever flaws or relevant suggestions expressed in percentage that is compared to maximum potential percentage.

66
 O.Mungunbayar students from Baganuur districts’
participating schools.

B. Following set of recommendations were made for the

program developers based on feedbacks of participants
of the experiment.
x Make the registration part shorter
x Regularly update and fix identified flaws with
program writing
x Look at similar programs, internationally
x Update website and interface of the program as
well as make them more tailored to children

Fig. 2. Initial scores across the abovementioned 3 groups

CONCLUSION
In 2018, “MATHSHOP” program was initially
introduced in 2 kindergarten and 2 secondary schools
(schools # 33 and “SHINE UE”) in order to demonstrate
programs upbringings and positive results were disseminated
at the number of events. This time, we supported high school
students from 4 secondary schools out of Baganuur district of
Ulaanbaatar city with their quest of preparing for the General
admission test and the collaboration was successful. Once
again, it is proven that “MATHSHOP” is the effective e-
learning program as overall test scores of our participants
improved, significantly.
The “MATHSHOP” program, which was built by’
leading scientists and professionals’ team, possesses crucial
advantages of meeting the modern innovative teaching
techniques and methodologies, stimulating students’ self-
Fig. 3. Comparison of scaled scores between national and Baganuur district learning skills while being cost-effective and pandemic
values from the last 2 years (2019, 2020) sensitive. Thus, we extend our invitation for collaboration to
decision makers of the education sector, entrepreneurs,
TABLE.I THE COMPARISON OF SCALED SCORES BETWEEN program developers and all other interested parties, in scaling
NATIONAL AND BAGANUUR DISTRICT VALUES FROM THE LAST up the “MATHSHOP” e-learning program’s use at the
2 YEARS (2019, 2020)
national level.
The national Average of Baganuur
Year average district
REFERENCES
2019 495.05 502.6
2020 500.01 524.99
Improved [1] P.Lkhagvasuren Ph.D, U.Oyunkhand, D. Erdenebat, B.Otgontugs,
percentage of Interpreting student’s knowledge in Mathshop program
scaled scores 1.0 4.5 implementation of Bayesian knowledge tracing, Management and
Innovation scientific journal, №2(008), pp.103-107, 2018
[2] P.Lkhagvasuren Ph.D, U.Oyunkhand, D. Erdenebat, B.Otgontugs,
By clicking on the help button we could always find Interpreting student’s knowledge in Mathshop program
necessary support without going through books. As a result implementation of Bayesian knowledge tracing, IBS International
of participating in this experiment, our self-learning skills Symposium on Computer Science, Computer Engineering and
Educational Technology 2019, Wolfram Hardt, pp.44-48, 2019
and methodologies have gotten better. Most importantly, we
[3] П.Лхагвасүрэн, Д.Эрдэнэбат, Б.Ганцэцэг, Математикийн
got accepted to the schools we wanted, thanks to much хичээлийг цахимцуулах нь, МУ-ын ШУТИС Эрдэм
improved General admission test score. Last but not least, we шинжилгээний бүтээлийн эмхтгэл, pp20-24, 2018
would like to thank the MathShop team for helping us out [4] http://eec.mn/?p=956
and inviting our feedback when developing this great [5] http://list.eec.mn/
program, further” says M.Tsenguun, E.Khosbayar and

67
 Collision Warning Based on Multi-Object
Detection and Distance Estimation

Shadi Saleh Chirag Rellan Shubham Pareshkumar Surana

Department of Software Engineering Masters Embedded Systems Masters Embedded Systems
Technische Universität Chemnitz Technische Universität Chemnitz Technische Universität Chemnitz
Chemnitz, Germany Chemnitz, Germany Chemnitz, Germany
[email protected] [email protected] [email protected]

Julkar Nine Wolfram Hardt

Department of Computer Engineering Department of Computer Engineering
Technische Universität Chemnitz Technische Universität Chemnitz
Chemnitz, Germany Chemnitz, Germany
[email protected] [email protected]

Abstract— Road accidents are a leading cause of non-natural close. Integrating Distance Estimation with Object Detection
deaths around the world. Traffic safety is of utmost priority in could play a significant role in efficient Collision Avoidance
the modern age. This problem can be reduced by implementing Systems[4] and can support a wide range of applications [5].
a collision warning system that tells the driver when a vehicle is
too close. However, assisting the driver in real-time is a
However, this task has been extensively investigated using
challenging task. Traditionally, it requires high-end hardware traditional Computer Vision (CV) methods. Nowadays,
and different and sophisticated algorithms for different integrating Object Detection and Distance Estimation based
environments. These algorithms require high expertise, to on visual sensors have gained much attraction and
create and deploy inside a vehicle. In this paper, we discuss deep outstanding outcomes since the introduction of Deep
learning-based approaches to solve the problem at hand. Deep Learning [6]. A Deep Learning approach has proven its
Neural Networks are notorious for their high computational
costs and latency while performing object detection. Two
outstanding performance in high-level perceptual and
approaches with low inference time and reduced complexity are cognitive tasks such as detection, recognition, and
discussed here to create a setup to perform real-time object understanding of scenery. Neural networks are much more
detection with acceptable accuracy. A system is created such suited to such tasks than traditional CV methods, as they [7]:
that a camera can be mounted onto a car that captures the scene x Have higher accuracy when used for tasks such as image
in front of it. Then the Deep Neural Network detects and
classification and object recognition.
classifies the objects in the scene and estimates their distance
from the host vehicle. A warning is then displayed when the x Require less problem-specific expertise.
distance is low to avoid a collision. The warning can be in the x Provide more flexibility: the same neural network could be
form of a visual cue, to alert the driver in advance. retrained for a different problem.
Deep Neural Networks such as RCNNs[8], YOLO[2], and
Index Terms—, DNN- Deep Neural Network, SSD- Single Shot SSD MultiBox[8] can be deployed for the vehicle detection
Detector, TensorRT, YOLO- You Only Look Once.
task.
I. INTRODUCTION The remainder of this study is organized as follows: in the
next section stating the goals of the study. In section III,
Road traffic accidents, otherwise known as, “The
former work concerning object detection and related fields is
Neglected Epidemic” are a leading cause of non-natural
reviewed. Gaps in those works are examined briefly here.
deaths around the world. According to WHO, 20-50 million
Next, the methodology of our proposed methods is shortly
people suffer non-fatal injuries every year in traffic accidents
described in section IV. Afterward, in section V, a
[1]. Vehicle-to-vehicle crashes are the most common
comparison between existing approaches and our approach
amongst road traffic mishaps. Smart vehicle systems such as
alongside results is provided. Section VI concludes the paper
Vehicle detection and Collision warning can be used to
and states the next steps of the proposed research.
improve traffic safety. Humans need only glance once and
immediately know what is in front of them, how they behave II. RESEARCH OBJECTIVE
[2]. Computers and automation systems, on the other hand,
Our main objective is to predict to estimate a light
do not perceive the surrounding environment in the same way
paradigm/model of object detection and distance estimation
as humans do. “Object Detection” is spotting an object of
with minimal space and power consumption requirements as
interest in the scene, then, localizing it by drawing a bounding
well as real-time capability. Nowadays, a variety of neural
box around it and classifying it into one of many different
network architectures are available for object detection.
classes [3] e.g. Human, Car, or Bus, etc. “Distance
However, due to different performance parameters such as
Estimation” is approximating the distance between the
inference time, mAP (mean Average Precision), etc. not
detected object, normally another vehicle or target object, and
every neural architecture can be employed for our purpose. In
the host vehicle. “Collision Warning” is alerting the driver of
this paper, we investigate models that are appropriate for
an impending collision in advance if the detected object is too
inference on embedded GPUs.
68
A. Why is this important? detection without specific feature definition. They are based
Deep Neural Networks are computationally expensive. on the Convolutional Neural Network (CNN). There are
This is primarily because millions of parameters need to be mainly 2 types of deep learning networks, i.e. region proposal
calculated and tuning for performing inference on even a networks and 1-level networks [13]:
single frame of a video. Thus, high-performance hardware, Region proposal networks: require 2 stages, one for
such as a desktop-grade computer complete with a discrete generating region proposals, and the next to detect objects in
graphics card, is required to perform object detection (or the proposed region. They include networks like RCNNs and
inference) in real-time. However, the hardware available in a its successors.
vehicle has only limited resources. A solution must, therefore, On the other hand, Single-stage networks include networks
be developed to reduce the computational requirements of like YOLO and SSD. They only require 1 stage/ a single
these DNNs. forward pass of the network to detect multiple objects in an
In this paper, 2 different DNN architectures are examined, image/frame from a video.
namely "Tiny YOLO" and "SSD MobileNet", which are Single-stage networks like YOLO and SSD are faster than
tailor-made for low-performance hardware. These are their region proposal counterparts but have lower accuracy
implemented on a development board with embedded GPU and mean Average Precision (mAP) [14]. Additionally, they
to emulate the type of hardware inside a vehicle. The board have some problems detecting smaller objects. From the table
used here is NVIDIA Jetson Nano. A monocular camera is above, single-stage networks have significantly higher
utilized to capture the scene in real-time and provide input to inference speed (FPS) while not much trade-off in terms of
the models. accuracy (mAP).

Table 1 STATE OF THE ART TECHNIQUES FOR OBJECT

IV. METHODOLOGY
DETECTION In this section, the proposed method for carrying out the
research is discussed.
Benchmark Frame mAP
Technique Year Verdict
used rate (%)
Feature Images Works with
2004 - ~97%
Extraction [9] from roads images
GMM based Static camera
2016 Real life 30 96.85%
[10] position
Faster R-CNN Pascal VOC
2015 7 73.2 Low fps
(VGG-16) [2] 2007
Pascal VOC Good overall
SSD-300 [11] 2016 46 74.3
2007 performance
Pascal VOC Low fps at this
SSD-500 [8] 2016 19 76.8
2007 resolution
Better
Pascal VOC performance
YOLO [2] 2016 45 63.4
2007 than fast
YOLO Fig. 1 General workflow of a DNN
Very high
Pascal VOC
Fast YOLO [2] 2016 155 52.6 frame rate but
2007
low precision I. Models
a. Tiny YOLOv3: Tiny YOLOv3 is a simplified version of
III. LITERATURE REVIEW YOLOv3, which is a state-of-the-art Object Detection
Nowadays three approaches are possible for object System. The only difference is that there is a lesser number
detection: Traditional CV techniques, approaches based on of convolutional layers in the Tiny version than YOLOv3. It
Machine Learning, and approaches based on Deep Learning. requires less computational power than full-scale YOLOv3.
Table 1 above shows a comparison of the techniques from Hence, it is ideal for deployment onto Jetson Nano with
different approaches. restricted computing capacity. The Fig.2 below shows the
architecture of Tiny YOLOv3.
Traditional Computer vision approaches, which work on
hand-crafted feature definition [7]. However, roads are
dynamic, meaning that there can be different illuminations,
backgrounds, etc. The location of vehicles, their size, and
shape can be everchanging.
Machine learning models provide a better solution to this
problem. They can extract features from input data and
perform classification [12]. However, domain expertise is
Fig. 2 General Architecture of Tiny YOLOv3 [15]
required to design feature extractors and the features are not
scalable. b.MobileNet-SSD-V2: MobileNet-SSD-V2 was deployed on
Deep learning models, these models are more suitable for Jetson Nano and inferencing was done using an optimized
automotive applications as there is a lot more compute power TensorRT engine as when combined they help reduce
available nowadays than was previously available. complexity and speed up the inference.
Additionally, there is a lot more data e.g. ImageNet dataset
that can be used to train a deep neural network. Neural
networks inherently have the advantage of being retrainable
Deep learning approaches can perform complete object

69
 Fig. 4 MobileNet V2 Layers as used in MobileNet-SSD-V2

Optimized TensorRT engine Inference on Jetson Nano uses

Fig. 3 General Architecture of MobileNet-SSD-V2 [16] an optimized TensorRT Engine. It increases throughput and
lowers latency while inferencing. It offers several precision
SSD- Single Shot Detector is a technique developed by optimizations such as INT8 and FP16[20]. This decrease in
Google research teams. It only requires a single forward pass inference precision translates in many cases to reduced
of the network to perform classification as well as localization latency without sacrificing accuracy much. Additionally,
on an image/video frame. It is designed to be used larger models can be made to fit in memory. Thus, with the
independently of the base network. MobileNet architecture is help of these powerful optimizations, MobileNet-SSD-V2
used as the feature extractor for the SSD. The Fig. 3 above can perform exceptionally on low-end hardware such as
shows the architecture of MobileNet-SSD-V2. Jetson Nano.
II. Training and Dataset
In order to implement our proposed approach, it is very
important to use an adequate dataset for the training with
relevant traffic data. Different datasets are available for object
detection. The two datasets which are used for training the
above-mentioned models are listed below:
KITTI Dataset: KITTI dataset is used to train Tiny
YOLOv3. It contains images labeled with eight different
classes for a total of 7481 images for training [17]. Mostly,
the images are focused on 2 classes namely cars and
pedestrians. Classes vans and sitting persons are not counted
as false positives for cars and pedestrians respectively due to Fig. 5 Inference process TensorRT engine[20]
similarity in appearance. The proposed model was trained on
IV. Distance Estimation
a PC with Intel Core i7 9th Gen processor, NVIDIA GTX
Distance estimation is performed alongside vehicle
1660 TI GPU, 16 GB RAM. CUDA10 and CUDNN7.3 GPU
detection on Jetson Nano. However, an approximation of real
accelerated libraries were used. The training was done on the
distance is unreliable due to the use of a monocular camera,
Darknet framework using an 80-20 split of KITTI Dataset.
inaccuracy in bounding box prediction, etc. Alternatively, a
The training was done until 250000 iterations with an average
relative method is used to estimate distance using the
loss of 1.1. The weight file with the highest mAP and lowest
bounding box dimensions and the total frame size (field of
loss was selected for deployment purposes.
view). The relation is as described below:
COCO Dataset: MobileNet-SSD-V2 was pretrained on
COCO Dataset- Common Objects in Context. COCO has 91
different object classes with more than 100,000 images for
training [18]. It contains all the necessary classes required for
Where W= ratio of the width of the predicted bounding box to
automotive purposes. But the model was modified to detect frame width (between 0 and 1).
only 6 classes relevant to the purpose of this research
including Bus, Cars, Pedestrians. When the distance calculated is less than 0.5, then a warning
III. Optimizations is displayed meaning that the object is too close! As the
The optimizations that were made in the two architectures objects get closer, the distance reduces quadratically. The
that have been employed in this research are listed below. relation is shown in Fig. 6.
Tiny YOLOv3: Tiny YOLOv3 uses the pooling layer instead
of YOLOv3’s convolutional layer with a step size of 2 to
achieve dimensionality reduction.
MobileNet-SSD-V2: MobileNet Feature Extractor:
MobileNet layers were used in SSD, to reduce processing
requirements by attacking the convolutional layers. It
converts traditional convolutions into depth-wise separable
convolutions, thus speeding up calculations while having a
very little accuracy trade-off [19]. The Fig.4 below shows the
general MobileNet-V2 approach.

Fig. 6 Relation between Width of the bounding box and frame width

70
 SSD
91 Jetson Nano +
MobileNet 28-30 Poor detection
(COCO) Webcam
v1
Jetson Nano +
Pednet 1 9-10 Poor detection
Webcam
91 Poor Detections
YOLOv3 Jetson Nano 4-5
(COCO) and very slow
Tiny
9
YOLOv3 Laptop 32-35 Good Detection
(KITTI)
on PC

Fig. 7 Outputs from both networks running on Jetson Nano Hardware

Fig. 8 MobileNet-SSD-V2 Output on a still image

Fig. 10 FPS/Wattage metric compared with different models on Jetson

Nano as well as PC

Fig. 9 Tiny YOLOv3 output on a still image

V. EVALUATION
In the Fig.7 are the screenshots of results from models
MobileNet-SSD-V2 (a) and Tiny YOLOv3 (b) respectively
running on the same video with the resolution 640 x 480 on
Jetson Nano. Tiny YOLOv3 was able to achieve 18-19
during the run of the video. MobileNet-SSD-V2 was able to
achieve 24-25 FPS on the same video. The differences were
primarily due to TensorRT’s memory optimizations. Both
models had very similar performance otherwise. In Fig.8 and
9 are the outputs of MobileNet-SSD-V2 and Tiny YOLOv3
on an image. Both the original images are provided for
reference on the left. In Table 2 below, are the results from Fig. 11 FPS/Cost metric compared with different models on Jetson Nano as
different models run on Jetson Nano as well as on a laptop. well as PC
The laptop used to run the test was equipped with Core i7-
Fig.10 and Fig. 11 illustrate the relationship between the
9750H CPU, 16GB DDR4-2666 RAM, and NVIDIA
FPS/Wattage and FPS/Cost metrics which can be used to
GeForce GTX 1660 Ti GPU with 6GB GDDR6 VRAM. For
compare the performance of the models used in this study and
consistency, most of the tests were performed on live webcam
other candidate models for a similar purpose. The higher
at the resolution 640 x 480.
these metrics, the better the model is on an embedded device
Table 2 COMPARISON OF DIFFERENT TECHNIQUE ON JETSON like Jetson Nano. Here, it can be seen SSD MobileNet v1 has
NANO AS WELL AS PC the highest FPS/Cost as well as FPS/Wattage metrics.
However, detections were poor. For a good balance,
No. of Deployment MobileNet-SSD-V2 and Tiny YOLOv3 were chosen. They
Model FPS Performance
classes Hardware
Good detections had good detections without compromising much in terms of
SSD
7 Jetson Nano +
23-25
but worse than inference times or FPS (See Table 2). YOLOv3 on Jetson
MobileNet
(COCO) Webcam SSD Inception Nano had very poor detections and slow inference time
v2
v2
Slightly better
around 250 milliseconds, thus making it unusable for real-
Tiny 9 Jetson Nano + detections than time applications.
19-20
YOLOv3 (KITTI) Webcam SSD MobileNet Thus, the models in this research outperformed the full-
v2
scale YOLOv3 on Jetson Nano by about 5 times in
SSD Very good and
91 Jetson Nano + FPS/Wattage Metric and 3.5 times in FPS/Cost metric.
Inception 18-20 precise
(COCO) Webcam
v2 detections Additionally, Tiny YOLOv3 on PC is also proven to be a lot
less efficient than when deployed on Jetson Nano. It offers
similar performance on limited hardware as well.
71
 VI. CONCLUSION AND FUTURE SCOPES [6] S. Saleh, S. Manoharan, and W. Hardt, “Real-time 3D Perception of
Scenes with Monocular Camera”, Embedded Self Organizing Systems
Technology has come a long way. Compact devices and (Vol 7. No 2. 2020). (accepted).
edge computing hardware are getting more powerful and [7] N. O’Mahony, S. Campbell, A. Carvalho, S. Harapanahalli, G. V.
powerful devices are getting more compact. Driver assistance Hernandez, L. Krpalkova, et al., "Deep learning vs. traditional
computer vision", Proc. Sci. Inf. Conf., pp. 128-144, 2019.
with the help of vehicle detection in real-time is no longer a
[8] Liu, W., Anguelov, D., Erhan, D., Szegedy, C., Reed, S., Fu, C.Y. and
distant dream. Notably, the complete setup used in the study Berg, A.C., 2016, October. SSD: Single shot multibox detector.
for implementation is a Jetson Nano and a CSI Camera, which In European conference on computer vision (pp. 21-37). Springer,
in total costs about 150€! Besides, it only requires a 20W Cham.
[9] Huang, S.S., Chen, C.J., Hsiao, P.Y. and Fu, L.C., 2004, April. On-
power supply to perform the complete calculations in real- board vision system for lane recognition and front-vehicle detection to
time. enhance driver's awareness. In IEEE International Conference on
Overall, it can be concluded that nowadays, there is no Robotics and Automation, 2004. Proceedings. ICRA'04. 2004 (Vol. 3,
pp. 2456-2461). IEEE.
need for more expensive and powerful hardware for such
[10] Seenouvong, N., Watchareeruetai, U., Nuthong, C., Khongsomboon,
tasks. These improvements could and will certainly be K. and Ohnishi, N., 2016, February. A computer vision-based vehicle
successfully implemented to improve road safety and ADAS detection and counting system. In 2016 8th International Conference
applications. A good example is a robust collision warning on Knowledge and Smart Technology (KST) (pp. 224-227). IEEE.
system based on the detection of multiple objects and the [11] J. Redmon and A. Farhadi, YOLO9000: Better faster stronger, 2016,
estimation of distance, which is presented in this study. [12] Singh, K.B. and Arat, M.A, “Deep Learning in the Automotive
Industry: Recent Advances and Application Examples,” arXiv preprint
As part of the future aspect of this study, a precise approach arXiv:1906.08834, 2019.
for collision warning should include more features and [13] Z. Zhao, P. Zheng, S. Xu and X. Wu, "Object Detection With Deep
additional optimizations, such as the integration of depth Learning: A Review," in IEEE Transactions on Neural Networks and
Learning Systems, vol. 30, no. 11, pp. 3212-3232, Nov. 2019, doi:
estimation with the object detection architecture. 10.1109/TNNLS.2018.2876865.
Furthermore, using the TensorRT engine with Tiny YOLOv3 [14] P. Adarsh, P. Rathi and M. Kumar, "YOLO v3-Tiny: Object Detection
will improve performance and allow better inference to be and Recognition using one stage improved model," 2020 6th
achieved. However, this should be considered and optimized International Conference on Advanced Computing and
Communication Systems (ICACCS), Coimbatore, India, 2020, pp.
in future work. Also, Hybrid Data processing: To improve 687-694, doi: 10.1109/ICACCS48705.2020.9074315.
detection in complex environments, combining and [15] D. Xiao, F. Shan, Z. Li, B. T. Le, X. Liu and X. Li, "A Target Detection
processing data from different sensors like LIDAR, IR, etc. Model Based on Improved Tiny-Yolov3 Under the Environment of
can prove to be useful. [21]. Mining Truck," in IEEE Access, vol. 7, pp. 123757-123764, 2019, doi:
10.1109/ACCESS.2019.2928603.
REFERENCES [16] Arabi, S., Haghighat, A. and Sharma, A. (2019). A deep learning based
solution for construction equipment detection: from development to
[1] Global status report on road safety 2018. Geneva: World Health deployment. arXiv preprint arXiv:1904.09021.
Organization; 2018. License: CC BYNC-SA 3.0 IGO.
[17] Geiger, A., Lenz, P., Stiller, C. and Urtasun, R., “Vision meets
[2] J. Redmon, S. Divvala, R. Girshick and A. Farhadi, "You Only Look Robotics: The KITTI Dataset,” International Journal of Robotics
Once: Unified, Real-Time Object Detection," 2016 IEEE Conference Research (IJRR), 2013.
on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV,
[18] Lin, T.Y., Maire, M., Belongie, S., Hays, J., Perona, P., Ramanan, D.,
2016, pp. 779-788, doi: 10.1109/CVPR.2016.91.
Dollár, P. and Zitnick, C.L., “Microsoft coco: Common objects in
[3] Phillips, Derek J., et al. "Real-time Prediction of Automotive Collision context,” European conference on computer vision, pp. 740--755,
Risk from Monocular Video." arXiv preprint 2014.
arXiv:1902.01293 (2019).
[19] Howard, A.G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W.,
[4] Saleh, S., Khwandah, S., Heller, A., Mumtaz, A. and Hardt, W. (2019). Weyand, T., Andreetto, M. and Adam, H., “Mobilenets: Efficient
Traffic Signs Recognition and Distance Estimation Using a Monocular convolutional neural networks for mobile vision applications,” arXiv
Camera. In: 6th International Conference Actual Problems of System preprint arXiv:1704.04861, 2017.
and Software Engineering. [online] Moscow: IEEE, pp.407-418.
[20] “NVIDIA TensorRT: Programmable Inference Accelerator,” NVIDIA,
Available at: http://ceur-ws.org/Vol-2514/.
[Online]. Available: https://developer.nvidia.com/tensorrt. [Accessed
[5] Saleh, S., Hadi, S., M., Amin Nazari, and Hardt, W. (2019). Outdoor 10 08 2020].
Navigation for Visually Impaired Based on Deep Learning. In: 6th
[21] T. Gandhi and M. M. Trivedi, "Pedestrian collision avoidance systems:
International Conference Actual Problems of System and Software
a survey of computer vision based recent studies," 2006 IEEE
Engineering. [online] Moscow: IEEE, pp.397-406. Available at:
Intelligent Transportation Systems Conference, Toronto, Ont., 2006,
http://ceur-ws.org/Vol-2514/.
pp. 976-981, doi: 10.1109/ITSC.2006.1706871.

72
 The Study to Determine the Required Product
Identification System
Sod-Od Batzorig Gundsambuu Bold
Department of Computer Science, Department of Computer Science,
School of Information and Communication Technology, School of Information and Communication Technology,
Mongolian University of Science and Technology Mongolian University of Science and Technology
Ulaanbaatar, Mongolia Ulaanbaatar, Mongolia
[email protected] [email protected]

Abstract— Demand and publication of commodities is a key 4. Organizations' information technology maintenance and
factor in the trade sector. It is possible to predict random demand assistance costs will decrease.
from consumers and find an optimal amount of resources.
Therefore, we examine the methodology for estimating the optimal 5. When large surveys and calculations are needed which
quantity of resources using data which recorded on the database required to estimate for a few years on the average
in the inventory of the commodity (stock) demand. The cloud computer will be able to run fast enough for multiple
computing will be used to estimate the optimal quantify and servers to compile.
calculation section uses mathematical techniques.

Keywords –resource depletion and surpluses; optimal quantity of

resource; computing software; data processing; mathematical
methodology;

I. INTRODUCTION
It is a good idea to implement the programming and
technology solutions for development of the system is based on
cloud-based technology solutions. With the rapidly updating of
information technology around the world, the latest information
technology has been developing in our country. An example of
this is cloud computing technology and its products. Cloud
computing is a technology development has brought big changes
to the information technology industry. Fig. 1. Cloud computing [3]

As a programming solution, we use PHP which is the world's For this development system, cloud computing is based web
most widely used programming language [1] [2] and MySQL platforms [4] [5] and also allow to use their mobile phones to
for databases. Developing a program based on cloud-based manage their full-featured system. From all this, there are many
solutions provides the following advantages. advantages.
1. The user will be able to access their information and II. RESEARCH
applications wherever they are and it is enough to
connect to the Internet. The information is not available If the resource amount is random, its demand is a random
on the user's computer and is not restricted to the user's quantity. The following values are used to find the optimal
computer memory. quantities of demand for randomly generated demand [6]. These
includes:
2. The cost of hardware of organization which is receiving
services is reduced. Because the high-performance x Real and practical quantities of demand (these values are
computers will be running the service provider side. regarded as registered in a database.);
Therefore, the user who receive the cloud service is x If the required resource exceeds the demand, the cost of
enough that will have average performance computer to unit surplus is c1;
connect to the Cloud system.
x If the required resources less than demand, the cost per
3. Some organizations are temporarily borrowed data
unit depletion is c2;
storage memory from other organizations because they
lack memory. Therefore, they will not have to worry The demand for resources is a random quantity ∈
about data storage memory problems. { , , ⋯ , , ⋯ } and if the resource is x then total cost will
be

73
 ( )= ∑ ( − ) ( )+ ∑ ( − ) ( ) (1) program getStartingValues
Do i = 0 to k + 1
Here, ( ) is probability that the demand for . The minimum
point of the function ( ) is the optimal amount of resource. set Ti = Xmin - h/2 + i*h

( ) is distribution function of random variable [7]. Then, end do

the optimal amount of resources would be the following. 3. Find optimal quantities:
program getCalculateResult
( )< < ( ) (2)
+ Do i = 1 to k
set Mi = 0
And the value of is to provide an inequality. The values of
( ) are calculated as statistical data of demand. Therefore, we Do j = 1 to n
use the realistic and practical data of the resource demand which If Ti < Xi and Xj <= Ti + 1
is calculated by setting the demand distribution function ( ),
to calculate the optimal amount of resource with the equation then
(2). set Mi = Mi + 1
We propose a hypothesis of what distribution is needed in end do
the statistics analysis of demand. This hypothesis will be verified
by statistical analysis [8]. In practice, the demand is often a setr = c2/(c1 + c2)
normal distribution. For example, if the demand has set P1 = M1/n
mathematical average, standard deviation, and its distribution
is normal, then the distribution function will be follows. Do i = 1 to k

1 1 − set Pi + 1 = Pi + (Mi+1)/n
( )= + ∙Ф (3)
2 2 if Pi < r and r <= Pi + 1
then
In the above equation, Ф( ) is a function of Laplace and the
function looks like this: set print Ti
2 /
end do
Ф( ) = (4)
√2 B. Implementation
Test data are shown below.
The optimal size of the product is the following equation and
calculate this equation, use the Laplace function table. TABLE I. EXAMPLES OF TEST DATA

Item code Date Quantity

1 1 −
+ ∙Ф = (5) 1 10/6/2017 0:00 9
2 2 +
1 10/6/2017 0:00 10
A. System pseudo code
1. Find the maximum and the lowest value of samples: 1 10/7/2017 0:00 19
1 10/4/2017 0:00 17
program getMinMax
1 10/5/2017 0:00 15
set Xmin = X1, Xmax = Xn
1 10/9/2017 0:00 14
Do i = 1 to n 1 10/8/2017 0:00 11
If Xi < Xmin 1 10/1/2017 0:00 12
set Xmin = Xi; 1 10/7/2017 0:00 15

If Xmax < Xi 1 10/7/2017 0:00 23

1 9/7/2017 0:00 22
set Xmax = Xi;
1 9/6/2017 0:00 5
end do
1 9/8/2017 0:00 1
set k = n/2
1 9/7/2017 0:00 19
set h = (Xmax - Xmin)/k
2. Find the starting values in the interval: x c1 or the cost of unit surplus is 250;

74
 x c2 or the cost per unit depletion is 50; The key issue of efficient resource management is
The following figure shows sales of goods and estimated x Resource accounting for inventory is need to determine
outcome is 5.7142857142857 (optimal amount of clearance). how much resources are available;
This result is an optimal amount of random demand for goods
clearance. In other words, it is to ensure that the surplus or x Based on the resource accounting, we can determine the
depletion of the goods is minimal. amount and timing of what resources to order or receive.
2. Inventory registration system
35 23 22 Inventory is the array of finished goods or goods used in
19
30 production held by a company [9]. Inventory is classified as a
current asset on a company's balance sheet, and it serves as a
25
19 17 15 15 buffer between manufacturing and order fulfillment.
20 14
12
11 5 1 x Assets held for sale during normal business operations;
15
10
10 x Items in the production stages for sale;
9
5 x Materials or supplies to be allocated or used for the
services;
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 Inventories constitute a major part of the working capital of
business and trade directives, and are the main source of
Fig. 2. Sales of goods business income. There are two systems for inventory
registering.
It is also possible to use the system to link the system with
the resource management then the data will add the database, it x Perpetual system - Each movement of the revenue and
can be automatically displayed at exactly one hour. This is expenditure of inventory recorded inventory accounts
especially suitable for organizations that hold permanent goods. and the inventory balance can be obtained from the
information in the registry. If the amount determined in
C. Inventory and resource system the inventory record is different from the amount of
In order to determine the optimal amount of product and physical count, make the correction record of the
production, the first thing we need is the notes of recorded on inventory.
inventory paper and data stored on the electronic records system. x Periodic system - It is impossible to determine the
1. The need for inventory to reserve balance of inventory from the information in the record
so the physical inventory will be made. There is no
There are several reasons why resources are reserved in the indication of a comparison with the physical count, so
industry. Below are listed a few important reasons. there is no surplus. Surplus and depletion are initially
x Back up stock to keep the production constant; contained in the cost of the goods, so correction is not
required.
x Backup to meet expected demand;
x Furthermore, the system can be incorporated into the
x Resources are stored to connect production, allocation enterprise resource planning system and it can be
systems and production phases. upgraded to software development.
x Protect against imbalance. Therefore, supply of
industrial resources does not break down or when it
comes to increased demand when resources are depleted.
x Use the order cycle advantages. Resources are reserved
for purchasing and subscription costs which are limited
to a certain amount.
x Sometimes resources are reserved for avoiding price
increases and probability of resource depletion is likely
to occur.
x Operations require a certain amount of time, and during
this period certain resources exist in the production
process.
The main objective of resource control is to provide
consumers with the desired demand, at the desired time and with
quality products that meet the needs and on the other hand to Fig. 3. An optimized system is in the ERP system [10]
minimize the cost of the process.

75
D. Screenshots of our development system In practice, the problem occurs in a widely that
Screenshots of software developed by the research work will determination of optimal quantity of resources with random
be demonstrated in the following. The figure 4 is the general demand. Based on the study of the state of demand for resources
display of our development system. The next figure shows a or demand distribution, the total cost is expressed by demand
section of the estimation. and other costs. Therefore, we consider the methodology for
finding the optimal value. The developed system can be fully
implemented in practice and implementation objects can be
retail stores and product manufacturers.
In the future, resource allocation is not uniform, in particular
when the expenditure is time-consuming, this research could be
expanded.
IV. REFERENCES
[1] R. Sandler, "The 14 most popular programming languages, according to
a study of 100,000 developers," Business Insider, 20 4 2018. [Online].
Available: https://www.businessinsider.com/14-most-popular-
programming-languages-stack-overflow-developer-survey-2018-4#8-c-
this-language-pronounced-c-sharp-was-developed-by-microsoft-its-a-
rival-to-the-even-more-popular-java-and-largely-used-by-business-
software-de.
Fig. 4. General view of our development system
[2] L. Welling and L. Thomson, PHP and MySQL Web Development.
[3] U. Marndi, "Disadvantages are addressed by cloud computing," Zeb Info
Tech, 22 6 2018. [Online]. Available:
https://www.google.com/url?sa=i&rct=j&q=&esrc=s&source=images&
cd=&ved=2ahUKEwjL9OyC0qDdAhUJfbwKHeD2CoQQjhx6BAgBE
AM&url=https%3A%2F%2Ffanyv88.com%3A443%2Fhttp%2Fwww.zebinfotech.com%2F2018%2F06%2F
22%2Fdisadvantages-are-addressed-by-cloud-
computing%2F&psig=AOvVaw2wgRCAv5iIULonvztXKk7B&.
[Accessed 15 8 2018].
[4] D. C. Marinesci, "Cloud Computing: Theory and Practice," p. 416, 2013.
[5] "Cloud Computing," Free Book Centre, [Online]. Available:
http://www.freebookcentre.net/Networking/Cloud-Computing-
Books.html.
[6] Takha, Introduction to Operations Research, 2005.
[7] "Distribution functions," PennState Eberly College of Science, 2018.
[Online]. Available:
https://onlinecourses.science.psu.edu/stat414/node/98/.
[8] "Sampling Distributions and Hypothesis Testing," Department of
Psychology University of Toronto, 2018. [Online]. Available:
Fig. 5. Estimation part of our development system http://www.psych.utoronto.ca/courses/c1/Baby_howl/Chap8/chap4b.htm
l.
III. CONCLUSION [9] "Inventory," Investopedia, [Online]. Available:
https://www.investopedia.com/terms/i/inventory.asp.
Within this research work, we have developed a [10] "Enterprise Resource Planning," IBI, [Online]. Available:
mathematical modeling methodology with programming and https://www.ibi.mn/news/201309091105265634.
technology and a cloud-based system to find the optimal amount
of product.

76
 Comparison of Photogrammetric Approaches for
Estimating Motion Parameters from a Monocular
Sequence
Tserennadmid Tumurbaatar
Department of Information and Computer Science
National University of Mongolia
Ulaanbaatar, Mongolia
[email protected]

Abstract—The techniques of measuring position and In this paper, we investigate methods for estimating 3D
orientation of the object using image correspondences are based motion of a moving object in image sequences taken by
on the principles of epipolar geometry in photogrammetric field. monocular camera in real-time. We have used a monocular
In this study, we aim to evaluate the main differences of web camera instead of other specialized magnetic devices to
approaches for estimating motion parameters, characterized by sense 3D motion since it is widely available. The aim of this
3D rotation and 3D translation. The estimation methods paper is analyze and compare the photogrammetric
developed in photogrammetric field were carried out. This approaches. We have used two estimation methods for
paper describes the mathematical motion model for the determining 3D motion parameters based on tracked feature
proposed approaches and comparison of their results, indicated
correspondences. They were developed in photogrammetric
the differences between the estimations in terms of accuracy and
test dataset. We showed that relative orientation based
approach. Firstly, we will introduce mathematical
approach produced accurate results for image sequences of 3D formulations for the proposed approaches. Secondly, we will
objects but homography based relative orientation approach explain implementation steps of 3D motion estimations
produced accurate results for image sequences of planar objects. regarding image sequences from single camera. Thirdly, we
will point out their practical differences in case of test datasets
Keywords—motion estimation, correspondence point, moving and analyze comparison results.
object
This paper is organized as follows. Mathematical
formulations of the motion models in photogrammetry will be
I. INTRODUCTION
described in Section 2. Methodology to implement 3D motion
Recovering position and orientation of the object from estimations and test datasets are presented in Section 3.
images is becoming important task in the fields of the Comparison results will be discussed in Section 4.
computer vision and photogrammetry. In photogrammetry,
the position and orientation (exterior orientation parameters) II. MOTION MODELS
between right image frame relative to left image frame in a We review the basic geometry of the proposed motion
sequence can be estimated by collinearity or coplanarity models. The relative motion of a moving object in image
equations [1]-[5] with a sufficient set of corresponding points. sequences is determined once the correspondence points have
In previous study, comparative study of motion estimations been established. We have assumed that a camera is
from image sequences in photogrammetry and computer stationary. Consider X ∈ ℝ object space coordinates of a
vision has been published, introducing their accuracy to point P on a rigid object moves to X ∈ ℝ object space
different datasets [6]-[8]. In computer vision, automatic coordinate of a point P with respect to a camera coordinate
relative orientation of image sequences has been widely system. Using a perspective projection model, the points
investigated with assumption of a calibrated camera. Relative P , i = 1,2 are projected at x ∈ ℝ image space coordinates
motion parameters of the perspective two views are computed of a point p at time t and at x ∈ ℝ image space
by decomposing an essential matrix through eight point coordinates of a point p at time t on the image plane as
correspondences in the image pair [9][10]. Essential matrix is shown in Fig. 1.
defined up to a scale factor for translations as set of linear
homogeneous equations by establishing feature
correspondences. The estimation methods of relative camera
motion from image point correspondences [11] and from less
than eight point correspondences [12]-[15] in various
applications such as visual servoing and control were
developed. Moreover, motion parameters of a camera relative
to a planar object can be estimated by decomposing a
homography matrix through point correspondences. The
numerical and analytical methods for 3D motion from
homography decomposition were introduced in detail [16].
The authors [17]-[22] also introduced efficient solutions to 3D
motion estimation based on homography matrix through point
correspondences for planar targets in augmented reality and
robot control applications. Moreover, an additional Fig. 1. Geometry of object in image sequences.
investigation of the photogrammetric techniques that estimate
the motion parameters.

77
 Due to the rigidity constraint of the object motion [23], Here, = [ , , ] is the unit normal vector to the 2D
P and P are related by rotation matrix R and translational plane , and is the distance from the plane to the optical
vector T: center of the camera as shown in Fig. 2. The five parameters
( , , , , ) of the relative motion and three parameters
x TRx = 0 (1) ( , , ) of the plane in object space can be estimated from
This is called coplanarity or epipolar constraint written in this equation by a non-linear least squares solution.
triple product of vectors from Eq. (1).
III. THE IMPLEMENTATION OF THE PROPOSED METHODS
A. Motion model from on relative orientation In this section, we explain implementation steps for
Non-linear equations in coplanarity condition can be performing the two estimations as shown in Fig. 4. Firstly, we
written as following: capture an initial frame at time and extract a template region
from it as shown in Fig. 3 (A). Then SIFT feature points as
yellow dots in Fig. 3 (B) are computed for the template. Here,
= , = (
we assume N extracted feature points that are automatically
tracked to the next each F frames. Secondly, when a new
where frame at time of a moving object is captured, feature points
are extracted for the new frame. After matching feature points
= cosφ ∗ cosκ;
of the new frame with the feature points of the template
= −cosφ ∗ sinκ; region, the best matches (corresponding points) are found by
= sinφ; a Brute Force matcher.
= sinω ∗ sinφ ∗ cosκ + cosω ∗ sinκ;
= −sinω ∗ sinφ ∗ sinκ + cosω ∗ cosκ;
= −sinω ∗ cosφ;
= −cosω ∗ sinφ ∗ cosκ + sinω ∗ sinκ;
= cosω ∗ sinφ ∗ sinκ + sinω ∗ cosκ;
= cosω ∗ cosφ;

Then triple-scalar product of the three vectors as Fig. 3. (A) Template region as light green rectangle extracted in a frame, (B)
Extracted features to be matched as yellow points in the template region.
following:
Before estimation of the motion parameters, outliers
∙( − )+ ∙( − )+ ∙( − among the best matches are eliminated by RANSAC [24]
) = 0. (3) method. Thirdly, when all processing steps of two consecutive
frames are accumulated as mentioned above, the proposed
Here, ω, φ, and κ are rotations about the axis, axis, estimations defined in the Section 2 are used to estimate
and axis, respectively. , , and are translations about motion parameters.
the axis, axis, and axis, respectively. The three
parameters of the rotation matrix R and the two components
of the translation as unknowns can be determined by a non-
linear least squares solution from Eq. (3).

Fig. 2. Geometry of planar object in image sequences.

B. Motion model from homography based relative

orientation
Since mapping between the right and the left image frame
is given by the homography, motion and structure parameters
can be obtained from a following non-linear equation.
Fig. 4. Implementation steps for performing the two estimations.
 x ≅ R + Tn x  (4)
IV. THE EXPERIMENTS OF THE PROPOSED METHOD
We implemented the proposed estimations in
photogrammetry with the C++ programming language, Visual

78
Studio programming tool and OpenCV 2.4.9 library on a PC,
Intel(R) Core(TM) i7-9700. We used a Microsoft LifeCam
with known calibration parameters. We experimentally

Fig. 5. Rotation of the object around axis for dataset of real scene.

compared results from the two proposed estimations by for both of the datasets. Particularly, PM_H produced more
preparing different test datasets such as real and simulated accurate results for image sequences of the moving planar
datasets. Video sequences for a moving object are captured object since planar pattern is dominating in test dataset. On the
with 640 × 360 pixel resolution. We used thousands of other hand, the motion parameters estimated from PM_RO are
image sequences of the different textured 3D and planar the most accurate than PM_H for dataset of 3D object with
objects to check estimation accuracy of the motion noisy point correspondences. But PM_H and PM_RO still
parameters. have produced comparable errors. Success rate of estimations
stay in range of large motion of the moving object.
For creation of the real dataset, we created video
sequences while translating object position along each axis up TABLE II. ERROR ANALYSIS FOR ROTATIONS AROUND , AND
to 200 mm and rotating around each axis up to 90 degrees in AXES FOR SIMULATED DATASET
the front of static camera. The number of feature point for real
Error comparison of simulated datasets for 3D object
dataset was around 150. One of the prepared datasets for real
/degrees/
scene of the moving object rotated in 0o-30o degrees around PM_H PM_RO
axis is shown in Fig. 5 as an example. Estimation of motion Max 1.647 1.589
parameters can be performed between an initial frame as Min 0.015 0.003
shown in the first frame of the Fig. 5 and the next frames as Mean 0.549 0.461
shown in the second to last frame of the Fig. 5.
For creation of the simulated dataset, 3D object was CONCLUSION
translated by up to 15 units along , and axes and rotated
by up to 90 degrees around , and axes in Euclidean In this study, we compared the approaches for estimating
space. The number of feature point for simulated dataset was 3D motion in photogrammetry with different datasets. We
around 25. The feature points from image sequences were checked estimation errors for the approaches by using
measured with 2-4 pixel noisy for the both dataset cases. corresponding points between template region and subsequent
frames. The results of estimations were accurate in change of
We notated the estimation methods for experiment as translation and rotation under noisy correspondences.
PM_RO and PM_H for relative orientation and homography Specially, the homography based approach was more accurate
based relative orientation, respectively. For these estimations, for estimating motion of the planar object, and relative
we checked accuracy of the estimated rotation parameters orientation based approach was more accurate for estimating
with comparing true (known) rotation parameters. For motion of the 3D object. This means that the accuracy of the
reference true value, we manually measured corresponding estimation approaches is depending on test dataset.
points from the image sequences, and estimated their precise
3D motion. We analyzed mean, maximum (Max) and ACKNOWLEDGMENT
minimum (Min) errors for results from two estimations for all The work in this paper was supported by the grant of
datasets as summarized in Table 1 and Table 2. National University of Mongolia (No. P2019-3714 ).
TABLE I. ERROR ANALYSIS FOR ROTATIONS AROUND , AND REFERENCES
AXES FOR REAL SCENE
[1] B. K. P. Horn, “Relative Orientation,” Int J Comput Vision, vol. 4, pp.
Error comparison of real Error comparison of real 59-78, 1990.
datasets for 3D object datasets for planar object [2] T. Schenk, “Digital Photogrammetry,” Terra Science: Laurelville,
/degrees/ /degrees/ USA, 1999.
PM_H PM_RO PM_H PM_RO [3] P. Wolf, B. DeWitt, B.E. Wilkinson, “Elements of Photogrammetry
Max 1.867 1.863 1.828 1.983 with Applications in GIS,” 4th ed. McGraw-Hill Science, USA, 2014.
Min 0.0004 0.0004 0.0001 0.0009 [4] J. Kim, H. Kim, T. Lee, T. Kim, “Photogrammetric Approach for
Mean 0.538 0.527 0.312 0.408 Precise Correction of Camera Misalignment for 3D Image
Generation.,” Proceedings of the IEEE International Conference on
Consumer Electronics, pp. 396–397, 2012.
As we see in Table 1 and Table 2 that the errors of the
rotation results are small and accurate for each two approaches

79
[5] J.C. McGlone, E.M. Mikhail, J. Bethel, “Manual of Photogrammetry,” [15] G. Chesi and K. Hashimoto, “Camera pose estimation from less than
5th ed. American Society of Photogrammetry and Remote Sensing, eigth points in visual servoing,” IEEE, International Conference on
USA, 2004. Robotics and Automation, 2004.
[6] T. Tumurbaatar, T. Kim, “Comparison of computer vision and [16] E. Malis and M. Vargas, “Deeper understanding of the homography
photogrammetric approaches for motion estimation of object in an decomposition for vision-based control,” Research Report 6303,
image sequence,” Proceedings of the IEEE International Conference INRIA, 2007.
on Image, Vision and Computing, Chongqing, China, 2018. [17] K. Kim, V. Lepetit, W. Woo, “Scalable real-time planar targets
[7] T. Tumurbaatar, T. Kim, “Comparative study of relative pose tracking for digilog books,” Vis Comput, vol. 26, pp. 1145–1154, 2010.
estimations from a monocular image sequence in computer vision and [18] H. Bazargani, O. Bilaniuk, R. Laganie`re, “A fast and robust
photogrammetry,” Sensors, 19(8), 2019. homography scheme for real-time planar target detection,” J Real-Time
[8] T. Tumurbaatar, D. Terbish, E. Adiya, “Computer vision approaches Image Proc, pp. 1-20, 2015.
for motion estimation of object in image sequences,” Proceedings of [19] Y. Mae, J. Choi, H. Takahashi, K. Ohara, T. Takubo, T. Arai,
International Conference on Frontiers of Information Technology, “Interoperable vision component for object detection and 3D pose
Applications and Tools (FITAT), Hong Kong, 2018. estimation for modularized robot control,” Mechatronics, vol. 21 pp.
[9] J. Weng, T.S. Huang, N. Ahuja, “Motion and Structure from Two 983-992, 2011.
Perspective Views: Algorithms, Error Analysis, and Error Estimation,” [20] E. Marchand, H. Uchiyama, F. Spindler, “Pose Estimation for
IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. Augmented Reality: A Hands-On Survey,” IEEE Transactions on
11(5), May 1989. Visualization and Computer Graphics, vol. 22(12), pp. 2633-2651,
[10] R. Hartley and A. Zisserman, “Multiple view in computer vision,” 2016.
Cambridge University Press, 2000. [21] B. Zhu, “Self-Recalibration of the Camera Pose via Homography,”
[11] G. Chesi, “Estimation of the camera pose from image point IEEE International Conference on Intelligent Computing and
correspondences through the essential matrix and convex Intelligent Systems, 2010.
optimization,” IEEE International Conference on Robotics and [22] C. Pirchheim and G. Reitmayr, “Homography-based planar mapping
Automation, 2009. and tracking for mobile phones,” 10th IEEE International Symposium
[12] K. Fathian and N.R. Gans, “A new approach for solving the five-point on Mixed and Augmented Reality, 2011.
relative pose problem for vision-based estimation and control,” IEEE [23] S. Ullman, “The Interpretation of Visual Motion,” Cambridge, MA:
American Control Conference, 2014. MIT Press, 1979.
[13] M. Sarkis, K. Diepold, K. Huper, ”A Fast and robust solution to the [24] A. Martin, M.A. Fischler, R.C. Bolles, “Random Sample Consensus:
five-point relative pose problem using Gauss-Newton optimization on A Paradigm for Model Fitting with Applications to Image Analysis and
a manifold,” IEEE International Conference on Acoustics, Speech and Automated Cartography,” Commun. ACM, pp. 381–395, 1981.
Signal Processing, 2007.
[14] D. Batra and B. Nabbe, M. Hebert, “An alternative formulation for five
point relative pose problem,” IEEE, Workshop on Motion and Video
Computing, 2007.

80
 Studying the Students Information
Technology Use
(On the example of the 2nd course of the Mongolian National Institute of Physical Education)

#1
Tuvshinjargal Radnaa #2Tserengavaa Tserenpurev
Doctor, Researcher MA,Teacher
Studies Programm Sector Theory and Methodology Department
Mongolian Institute for Educational Research Mongolian National Institute of Physical Education
Ulaanbaatar, Mongolia Ulaanbaatar, Mongolia
[email protected] [email protected]

Abstract-In today`s era of rapid development of data and demands or concepts related to the growing knowledge and
digital technology, it is vital for higher education students to skills for learning technology in the data and information
have knowledge and skills about how to learn and use the society we live in today.
information technology, learn to use the devices such as
computers and other hardware. Thus they will be able to work The program means intended for the training purposes
freely in the internet environment to obtain useful information. such as give certain level of computer knowledge to students,
learn to make the optimal decision in real perplexed
Keywords-The needs and use of students’ information conditions, work independently, especially obtaining skills to
technology. process data, control oneself, to correct one’s mistakes and get
the skills of proper use of the information received.
I. BACKGROUND Students who use the information will be able to get and
provide the information they need for their training, use it in
It is impossible to think of the current students separate
their lessons, assignments, tasks and homework, and watch
from the information technology. Therefore, it is much of an
movies and videos they need, in return, it helps to express
importance to equip them to work and live in the information
their views and opinions properly online or on social media.
technology society. The situation with information technology
infrastructure in major cities and rural areas including sums
and smaller units vary a lot. [1]. Therefore, the information Innovative aspects of this research work
technology & science subject has an important role to give the It identifies the level of use of information technology by
knowledge and skills of the information technology students at the School of Physical Education and Sports.
consumption.
Media equipments such as computers, audio, video, CD, Practical significance of the research
VCD, information, and communication networks, are used in As a result, the conclusions and recommendations of this
the process of students` developmet 'self-knowledge, research should have been included in the curriculum of sports
knowledge-based skills, and practical skills, taking into schools, and have their implementation be made a clear
account the specifics of students' mental and information contribution to the improvement of training activities, which
processing and perception. There is a real need to develop of practical importance for this research work.
effective methodologies for the information technology-based
professional training such as the proper use of the Internet,
websites, computer applications and Internet applications. II. STUDY RESULT
Process and results of the information technology
As of the first half of 2013, there had been 609,000
Facebook users registered in Mongolia, that was 22 percent of consumption
the population, but now, after 6 years the number has grown In this study, 36 males and 14 female or total 50 students
by 13 percent or it means that 35 percent of the Mongolian of the Mongolian National Institute of Physical Education
population are Facebook users. According to a survey, 70 participated and the study was made by 8 surveys in selected
percent of all users are young people aged between 18-35. [2]. methods.

In the United States and Spain, students and learners use The answer was given that 88% of the students answered
information technology on a computer to practice, strengthen “yes” and 12% of them said “no” when the question was
knowledge and skills, problem solving and for continuously asked from the students whether they are updated with the
improving their information management and learning skills. information technology enthusiastically. According to the
In many countries around the world, the know how on how to study, it shows that majority of the students receive the
incorporate the latest achievements of information technology information then processed it, interact with others and stay up
into the educational teaching process and the right selection of to date about the news in the society. Nowadays lifestyle
the technology are being continuously conducted through demands them to use the information technology based
various researches with combined experiments and studies. electronic devices and equipment in order to keep up with the
[3]. technology development. Therefore, the students cannot
imagine or limit the information technology only by phones
The use of information technology in education plays a and Facebook.
key role in the process of introducing and developing
information technology in the education sector. The use of When the question was asked “How does information
information technology in education is related to new technology development influence your life?“, 96% replied

81
that it affects them well. However, 4 % replied that it was not
in the level to affect. This is a huge indication that the students
understand that the skills to work on the program use they
need to obtain at the information technology & science subject 0% 12%
are indispensable necessity for today’s students.
4% Yahoo mail
According to the study, it shows that majority of the
2%
students use the Excel program which is designated to
process the table information and to do accounting using the Gmail
mathematics function. It shows that 8% answered that they
use the programs fully and that means the students don`t have Yahoo,
necessary possession of the programs evenly. Therefore, it is gmail
crutial to make the information technology & science subject Facebook
program more progressive [Fig.1] 82%
Others

1% Fig. 3. Which email services do you use?

22% To do the lesson,
26%
homework
To play It shows that the information technology development has
0%
developed rapidly according to the above chart. The reason is
To spend time the celular phones were only used to speak and sms messages,
2%
4% To watch videos
to get access to the internet, to get the information they must
and movies find, to make connection with the organizations, to get the
6% To get the news bank services such as making payments etc; time and all
and information human actions have been conducted through the devices.
to chat
Therefore it is clear that young students more likely to use the
To enter cellular phones. Tablet and Ipad etc; other equipment are
39% facebook used as well. [Fig.4]
Others

Fig. 1. In what purpose do you enter online electronic environment?

According to the study, it shows that most students get Ipad Phone
access to the online environment in order to get the news and Notebook 2% Others
2%
information. However 1% of the students answered that they 15% Tablet
do not havea access to internet, because of playing games or Stationary
other reasons. This means it indicates it is a good sign, computer Stationary
although it has been the not enough in this current data 7% computer
society. [Fig.2] Notebook

Tablet Ipad
2%
Others
12% Phone
8% 72%
Word
10% 25%
Excel Fig. 4. What information technology equipment do you use more?

PowerPoint Time is not spent in cyber environment. According to the

study it shows that the physical education students spend less
All
time than other students means that they are busy with their
45% Others training and educational activities. When the survey was
conducted among the regular university’s 242 students, 24%
answered that they get 1-3 hours cyber environment or use
Fig. 2. Which one of the computer usage programs do you use more?
internet. [4]. [Fig.5]
According to the study, most students communicate with
one another through Facebook sharing the information. It has
been observed that there are only a few students who use the
online email. [Fig.3]

82
 to make change to the subject program at the
university for information technology & science
4% subject.
16% From 1 to 3
¾ In the data era, when students want to study
successfully, it is essential to find the information
From 3 to 6 that is necessary for them as well as it is important to
18% be able to use and organize it, to build-up
Over 6 information, to process, to distribute and to evaluate
62% the electronic technology.
No ¾ This research work shows that we do not limit
ourselves to using electronic tools and applications
with only by computer lessons, but also we need to
improve our knowledge and skills by receiving tasks,
homework, getting news and information from
Fig. 5. How many hours a day do you get online electronic environment?
professional sports sites, and to use information
In this survey, through the information technology & technology properly in our daily lives.
science subject, most students obtain the knowledge on how to ¾ In the future, it will be an issue to face us to choose
use the information technological electronic means. It was the most proper ways to use e-learning materials, to
observed that the knowledge and skills to process the have a student-centered teaching methods that are
electronic means were increased. 2 percent answered that appropriate for the age-specific mentality, on the
there was no improvement about the information technology other hand it is important to improve the teaching
knowledge and it shows that the information technology & and learning methodologies in such ways that
science subject could not include and provide needs of some learners have all the opportunities to learn
students when the analysis was made. [Fig.6] independently.

BIBLIOGRAPHY
4% [1] Information and communications technology (ICT), “Information
2%
technology white book”,Ulaanbaatar, pp.185, 2014.(references)
Yes [2] D.Chimgee, J.Erdenechimeg,“Higher education”, Ulaanbaatar, pp. 301,
2018.(references)
No [3] J.Davaa, G.Sarantuya,“University teaching technology”, Ulaanbaatar,
pp. 102-103,2013.(references)
A little [4] D.Chimgee, J.Erdenechimeg,“Higher education”, Ulaanbaatar, pp. 303,
2018.(references)
[5] N.Jargalsaikhan, Ts.Tavkhai “Computer self-study book”.
Ulaanbaatar, 2012.
[6] Ts.Chimedlkham, B.Zolzaya,“Information science-1”, Ulaanbaatar,
2006.
94%
[7] L.Davaasvren, “The beggnner’s guide book to use computer”,
Ulaanbaatar, 2009.
[8] B.Zolzaya., “Computer”, Ulaanbaatar, 2011.
Fig. 6. Was there any improvement to your technology knowledge by [9] Ts.Tserengavaa, ”The importance of recognizing computer
having the computer lessons? specification”, Mongolian National Institute of Physical
Education,“Physical education, sports theory methodology” Magazine
Vol 2017-4 (58), Ulaanbaatar, pp. 37. 2017
III. CONCLUSION [10] D.Nasanjargal.,“Computer studies” manual. Ulaanbaatar, 2013.
[11] “Training manual to improve the Information science lecturers’
¾ The lessons and gains were analyzed and concluded profession”, Mongolian National University of Education, Ulaanbaatar,
on the basis of the summary for the study and 2005.
research results taken from the students about how to [12] www.google.mn
improve and use the information technology [13] www.msue.edu.mn
[14] www.hicheel.mn
knowledge, training on the use of electronic devices
and equipment with greater results. As a result, we
reached the conclusion that it would be appropriate

83
Blended Mentoring in Academic Internship Program
Ummay Ubaida Shegupta René Schmidt Martin Springwald
Department of Computer Engineering Department of Computer Engineering Department of Computer Engineering
Chemnitz University of Technology Chemnitz University of Technology Chemnitz University of Technology
Chemnitz, Germany Chemnitz, Germany Chemnitz, Germany
ummay- [email protected] [email protected]
[email protected] chemnitz.de chemnitz.de
chemnitz.de
Wolfram Hardt
Department of Computer Engineering
Chemnitz University of Technology
Chemnitz, Germany
[email protected]
chemnitz.de

Abstract— Mentoring is considered as one of the best teaching contrast to students' competence gain. For this reason, we are
methods based on the recent scientific evidences. Due to the investigating possibilities to extract feasible data from the
ongoing digitalization in the learning environment, the traditional LMS "OPAL". Furthermore, the state of the art of data
mentoring processes evolved to technology supported blended analytics using artificial intelligence methods needs to be
mentoring. In this paper, it is described how this blended
analyzed compared to standard rule based approaches
mentoring strategy has been integrated into the academic
internship program of computer engineering education with the providing the basis of data usage approaches.
objective of introducing students the scientific way of working.
The students work intensively on their own topic, which in turn is The mentoring objective is to bring students closer to the
a part of a complex system. Throughout the entire duration of the scientific way of working through close cooperation
internship, the students are supported by the mentors in order to supported by technical methods increasing the interactivity
maintain the scientific way of working and to achieve their project and intensifying the mentoring process. This report aims to
goals. Various technical systems are used to support the students explore a blended mentoring approach in context of a
ensuring a lively support flow. The results show the effectiveness workshop in combination with the video conferencing tool
of the method.
Big Blue Button [12] and the team communicator Slack [13]
Keywords—Blended mentoring, communication, virtual to ensure scientific working methodology during the project
support, learning management system, scientific methodology. work of the students.

I. INTRODUCTION II. THEORETICAL FRAMEWORK

Mentoring is one of the best didactic methods worldwide, A. Pedagogy of Mentoring
as evidenced by the recent research that include a mentoring Mentoring has already been proven effective to provide
component mainly in the fields of science, technology, support to university students to facilitate their learning and
engineering and mathematics (STEM) [1]. Furthermore, to provide guidance [6]. When situated within the context of
particular application-oriented disciplines of computer a complex real-world task like internship, mentoring has
science education are addressed in integrating mentoring to already been proven as one of the ways to foster the inner
provide learners the necessary support to develop mastery growth of the individuals [7]. The pedagogical key of
[2], [3]. The mentor uses lessons characterized by a rich academic mentoring is its use of modeling to help a student
understanding of the cognitive process and the mentee's explore the inner craft of the intellect [8].
individual needs and interests to generate a mutual mentoring
process [4]. On the other hand, internship as a form of In this academic internship program, a group of students
experiential learning gives higher education students has been supervised and facilitated based on the pedagogy of
valuable opportunities to discover the work-life firsthand and mentoring. All the three mentors have tried to illustrate and
to apply theoretical knowledge to practice [9]. transmit meta-cognitive strategies, techniques for scaffolding
the students’ own thinking and learning process. Mentoring
In order to explore mentoring approaches, mentoring focuses on relationship that is supported by the sociocultural
opportunities are created in the academic internship program. theory of human learning by Russian psychologist Lev
These are created by providing particular tasks for the Vygotsky [8]. Vygotsky emphasizes the social interaction in
students in the context of learning data extraction and the development of cognition [8]. An aspect of this theory is
processing. The large fraction of learning data is generated in that the potential cognitive development is limited to a “Zone
the learning management system (LMS) "OPAL" of TU of Proximal Development” (ZPD) [6]. The student learns by
Chemnitz. It provides the ability to access the data to participating in challenging activities with a “more competent
acknowledge student's learning experience and taking data other” who supports the elements of the work that the student
driven decisions to improve the quality of education. This is cannot do individually. Gradually the student takes over
known as learning analytics [5], which can give an in-depth direction and moves toward independence [8]. The
insight about the implemented mentoring approach with scaffolding techniques of the mentor helped them to get

84
gradual support of consultation, instructions, motivation The six students registered have chosen 5 of the following
along with providing continuous feedback to their learning topics. The voluntary chosen topics are:
activities [7]. Afterwards, the mentoring process steered the 1. Analysis and Implementation of Methods for iCal
students to take over the ownership of their own learning, as Calendar Data Extraction
it is supported by ZPD [8]. 2. Analysis and Implementation of Methods for ONYX
Test Evaluation
B. Blended Mentoring
3. Data Acquisition and Generation of Statistical Data
Upholding the great promise of traditional mentoring, it from Log Files
has been expanded to E-mentoring and blended mentoring 4. Implementation of a GUI for Data Visualization
with the evolution of E-learning and Web 3.0 technology 5. AI for Automatic Test Data Evaluation
[10]. Blended mentoring process has more than one means of Two students with individual solution strategies have chosen
communication including face-to-face mentoring sessions topic 4.
accompanied by communication or mediation through e-
learning tools [10]. Previously, mentoring used the benefit of B. Scaffolding through Blended Mentoring
email to maintain the virtual communication between the In this academic internship program, the higher order
mentor and mentee [1]. The development of various user- learning is facilitated via metacognitive scaffolding
friendly technologies used for communication and as techniques [9]. The phases of planning, monitoring and
teaching aid in e-learning has paved the way to practice and evaluation of this technique [9] were incorporated in the
apply the theoretical knowledge of mentoring [3]. applied blended mentoring approach.
In the planning and monitoring phases, the students got
As scaffolding in the ZPD was followed in this mentoring incremental mentoring support at the beginning of the
process, it had helped to easily realize the blended mentoring internship. Two face to face residential workshop session
support via e-learning tools. This enables the mentee to keep occurred which were respectively three days and five days
track of his or her respective implementation tasks with the long. The aim of these sessions were to provide intensive
minimum possible guidance from the mentors. mentoring support. The students completed the task of
understanding the requirements of the internship, state of the
In the above-mentioned framework, it indicates that the art analysis and developed their first plan of implementation
mentoring process combines traditional and online sessions; it in these sessions. Formal presentations of the mentees in
can be called blended mentoring approach accordingly [8]. these sessions were realized as formative evaluation,
This blended approach has enabled the mentor to provide task rendering the scope of continuous development in their
related continuous feedback and to facilitate the students to respective internship tasks.
sustain their motivation to complete their assigned tasks.
C. Research Objective In between, these sessions and the communication at the
later period of this internship between the mentor and the
Mentoring in internship is defined as the degree to which
students were maintained via Email and SLACK [13].
the supervisor or mentor assigns the challenging tasks to the
Whenever the students needed in-depth support, the online
mentees, and provides proper guidance in accomplishing the
meeting tool Big Blue Button [12] had been used.
tasks and purposefully helps to develop the mentee’s
cognition and skills for real life implication. Reasonably, the
Finally, in the summative evaluation phase, the last
mentees’ view of the applied blended mentoring approach is
official online workshop the students were able to show their
necessary to continue this support in the academic research
implementation individually by themselves. Here they were
internship program. Besides, the student’s task oriented goal,
expected as competent individuals, who could perform the
which will be described in the task description section, the
difficult tasks by their own as stated by the “Zone of Proximal
research objectives are to explore the mentees’ perception
Development”.
about the mentoring support, about their cognitive, affective
and skill development, their preferences of the mentoring C. Internship Task Description
modes, mentor’s roles and the effectiveness of different The objective to bring students closer to the scientific way
consultation media. Along with these, finding out the of working, therefore 6 Tasks in the context of learning data
shortcomings of the mentoring process to improve it in future extraction has been defined, based on working fields
is an underlying objective of this research. identified in cooperation with the ‘tech4comp’ project.
Reasoned by data protection law conditions, the person data
III. RESEARCH METHODOLOGY cannot be made available to students. For this reason an
A. Student Background evaluation course in OPAL has been created, which contains
all necessary course elements. The students can use it freely
There are six students registered to this academic
to generate all necessary data as examples for artificial data
internship. They are all Master students of Chemnitz
generation following the given data structure. This course has
university of Technology (TUC). They are mainly from two
mainly 6 types of elements: Registration-Element, Upload-
faculties – Faculty of Computer Science and Faculty of
Folder- Element for the files from supervisors, 6 types of
Electrical Engineering and Information Technology. The
Calendar-Elements, two types of Blog-Elements, five types
gender ratio can be considered as balanced with 4 men and 2
of Checklists-Elements and five types of ONYX Test-
women. Two of the students are enrolled in first semester
Elements. Log data can be acquired from all course elements
whereas the other students are enrolled within a range of 3 to
and these different types of elements provide different types
7 semesters.

85
of learning experience data of the students. For conducting conditions. On this basis, guidelines should be developed
learning analytics, these data needs to be exported from the enabling AI-based evaluation and feedback generation.
data archiving tool and exported function from the "Course 5) Data Acquisition and Generation of Statistical Data
evaluation" and "Statistics" functions of the course. The data from Log Files
are exported as XLS Files. To perform Learning analytics it For data analytics, it is necessary to automatically read the
is necessary to extract and evaluate the data providing the log data of the course providing the access time and usage
particular task basis. The Task description are as follow: frequency of different course elements and store it in a
1) Analysis and Implementation of Methods for iCal database. Therefore, the learning outcome of this task is that,
Calendar Data Extraction the intern will be able to develop a script or program which
Opal offers the possibility to integrate iCal calendars into reads all information of the exported data file, prepares it and
courses. For data analytics, it is necessary to automatically stores it in a database.
read the data of the calendar and store it in a database. In the first step, the data must be understood and the
Therefore, the learning outcome of this task is that, the intern information that can be extracted has to be identified. On this
will be able to develop a script or program, which reads all basis, the database design is to be jointly defined and
information of the calendar, prepares it and stores it in a implemented in cooperation with the other teams. The data
database. will then be automatically extracted and prepared for
In the first step, the data must be understood and the database storage and statistical analysis. Finally, the data will
information that can be extracted has to be identified. On this automatically be stored in the database.
basis, the database design is to be jointly defined and
D. Basics of Task Completion Strategy
implemented in cooperation with the other team members.
The data will then be automatically extracted and prepared The first step is to recognize the data and filter the
for database storage. Finally, the data will be automatically information. On this basis, the database should be defined and
stored in the database. implemented as team across all members and filled manually
2) Analysis and Implementation of Methods for ONYX with artificial data once. Afterwards, the data has to be
Test Evaluation automatically extracted and prepared for database storage and
Opal offers the possibility to integrate ONYX Tests into statistical evaluation. The exported data has to be visualized
courses. These tests can be extracted by the archive function in the fourth task using the manually data of the database for
from Opal as Excel files. For data analytics, it is necessary to the prototyping process. Subsequently, the data has to be
automatically read this data and store it in a database. presented as components of a website. For this purpose,
Therefore, the learning outcome of this task is that, the intern statistics on the temporal evolution and the current status of
will be able to develop a script or program, which reads all the semester have to be conceptualized in advance.
information of the exported data file, prepares it and stores it Knowledge queries are very important in a mixed learning
in a database. environment, but a lot of data has to be processed. One goal
In the first step, the data has to be analyzed as preparation of data processing is the individual feedback from the pupils.
for information extraction. On this basis, the database design In order to give good feedback, it is necessary to understand
is to be jointly defined and implemented in cooperation with the connection between tasks and draw conclusions about
the other team members. The data will then be automatically possible misunderstandings. Task five required that a
extracted and prepared for database storage. Finally, the data systematic literary survey has to be carried out, with clues for
will automatically be stored in the database. the analysis of concepts that serve to identify
3) Implementation of a GUI for Data Visualization misunderstandings. With the help of the results, it should be
The data exported from task 1 to 4 stored in the database possible to generate ratings and feedback with the help of an
are to be visualized in this task. Therefore, the learning AI.
outcome of this task is that, the intern will be able to design a
database together with the team members. E. Data Collection and Analysis
Afterwards, the data should be presented as web In this study, mentors overall experience has been
components on a website. For this purpose statistics have to captured via unstructured interviews. Whereas, a close ended
be created to visualize the temporal evolution (e.g. by questionnaire is used to gather data of mentees’ overall
semester) and the current state of the semester. experience of blended mentoring support in this academic
4) AI for Automatic Test Data Evaluation internship. Since, the sample size is small (Mentor =3;
Knowledge tests are very important in the blended Mentee/Intern =6), the data has been analyzed as qualitative
learning environment. However, a lot of data has to be data. Narrative analysis is conducted for the unstructured
processed. One goal of such data processing is the generation interview data. The questionnaire responses have been
of individual feedback for the respective pupil/student. To analyzed by the combination of thematic analysis and
create a good feedback it is necessary to understand the descriptive statistics.
connections between tasks and to be able to draw conclusions
about the actual misconceptions. At this point artificial IV. RESULTS & DISCUSSION
intelligence should be used to identify such misconceptions. A. Mentor’s Perspective
Therefore the learning outcome of this task is that, the intern The mentoring process conducted through the mentors
will be able to conduct a systematic literature survey (with has been based on the aspect of sociocultural theory of Zone
evidence), to analyse concepts for the identification of of Proximal Development (ZPD). The interns have been
misconceptions and, if necessary, to extract the required mentored intensively during the first stage of the internship.

86
According to ZPD, the scaffolding techniques have been used
during these intensive consultation sessions. The mentors PLW 33.33%
have tried to model similar examples of their assigned tasks
and answered all the possible queries of the mentees to clarify CS 33%
the confusions of the tasks. The mentees have got frequent CD 33%
direct constructive feedback for the first three months of their
internship. Then they have been working on the theoretical APS 67%
basis and conceptual development of their assigned task, as PS 50%
these are fundamental for the implementation of the technical
part. SC 67.00%
Overall, the mentors are satisfied with the guidance that 0.00% 20.00% 40.00% 60.00% 80.00%
they provided. All the mentee have been able to complete
their internship tasks. Since all the mentee have to works in Fig. 2. Mentee’s improvement via mentoring support with
[SC] Self Confidence
whole groups, it has been found out that all of them are [PS] Presentation skills
supportive to each other. They have been successfully [APS] Analytical/ problem solving skills
worked in the team because of their sound communication [CD] Career development
and social skills. As they have to finish their task with less [CS] Communication skills
[PLW] Personal attitude towards learning and working
possible guidance during the last months of the internship,
there has been a reduction in the role of ‘Instructor’ as
Figure 2 shows their affective and skill development in this
mentor.
whole mentoring process. As one of the mentee stated in the
In course of mentoring, the mentors have realized that
open ended response,
most of the mentees lack in programming experience resulted
in lower quality of the implementation part. The
“I was challenged to discover new horizons in a field I
implementation have been successful though the mentors had
initially knew nothing about, and in this process gained
higher expectation.
profound, non-trivial depth of advantageous knowledge. This
B. Mentee’s Perspective has helped me both on academic and career levels.”
The responses of the mentees from the close-ended
questionnaire about the mentoring support shows diversified This statement indicates the improvement in the self-
prospect about the provided mentoring process. All of them confidence as well as in their problem solving skills of the
reported about being mentored via face-to-face meetings, mentees. However, the mentees have realized the
online synchronous and asynchronous communication differentiated mentor’s roles according to their needs shown
media. in figure 3.

Trainer
BBB
Motivator

Email Networker

Emo. Supp.
SLACK
Instructor

F2F Consultant

Advisor
0.00% 20.00% 40.00% 60.00% 80.00% 100.00%
Fig. 1. Effectiveness of the blended mentoring media for face to face
Teacher
meetings (F2F), Workspace Chatroom Slack (SLACK), E-mail contacts
(Email) and Online meetings via BBB (BBB) 0% 20% 40% 60% 80% 100%
Fig. 3. Roles of Mentor
They have opined that the face-to-face meetings provided the
most effective mentoring support (See Fig.1). Among the Figure 3 above shows that the mentors have been realized
other modes of communication, Slack and online meeting mostly the role of advisor and motivator in all through the
room Big Blue Button have been reported as effective internship period. The other significant roles that the mentors
mentoring tool. Their responses show that they have been have played are consultant and teacher. The mentors have
most benefited by the mentoring support in increasing their also performed the role of networker, emotional supporter
self-confidence. Figure 2 shows their perception that their (Emo.Supp.) and instructor. The role that the mentor have
problem solving skill and presentation skill have been played the least is as the trainer. This supports the theory of
improved following their self-confidence. ZPD in rendering the scaffolding support via mentoring as
motivator and advisor and less as a trainer.

87
They have also been asked about their competency mentors. This finding indicate the role of mentors mostly as
achievement according to the Information Technology motivator and advisor that is in line with previous result.
Competence model (Fig.4).
Another important finding from the recommendation of
the mentees are that they have needed more support in their
technical skill development. This result is in line with the
above mentioned mentor’s perspective that the mentees have
less programming experience.

Above all, the findings of both the task related results and
mentoring questionnaire indicate that the mentees have
received adequate mentoring support. However, some
amendments are needed in the next round in terms of
considering the individual needs of the mentees. Otherwise,
the blended mentoring support has helped the mentees to
complete their assigned tasks.
V. CONCLUSION
In conclusion, the mentoring process has been a success
based on the results. Nonetheless, the insufficient
programming skills of the students, the mentors have been
able to help the students finishing their programming tasks.
The findings indicate that the mentees are introduced with the
scientific way of working in this mentoring process.
Moreover, the mentees’ have realized their inner
Fig. 4. Information Technology Competency Model [14] improvement in analytical thinking. Skills like this represent
the key competences essential to achieve in the field of
The reported responses show an average group-wise gain of
Computer Engineering.
‘Industry-Wide Technical Competencies’ 37.5%, whereas
Due to the small sample size, the findings can not be
the ‘Workplace’ and ‘Academic’ competencies are
generalized for all the internship program in computer
respectively 61% and 40.5%. Most of the mentee’s about
engineering. Even though, this process of blended mentoring
61.9% have reported their ‘Personal Effectiveness
has been successful reasoned by well-chosen technology.
Competencies’ gain. This result shows altogether the
cognitive, affective and skill development from mentee’s
point of view. In addition, it is according to our objective to ACKNOWLEDGMENT
bring the academic working style closer to the students.
The authors gratefully acknowledge the funding by the
BMBF (tech4comp 16DHB2106).
In this process of mentoring, the mentors’ characteristics
have been one of the crucial criteria that was asked to the REFERENCES
mentees. Among those, the following Tab. I shows the most [1] H. Stoeger, X. Duan, S. Schirner, T. Greindl, and A. Ziegler, “The
significant ones from mentee’s responses. effectiveness of a one-year online mentoring rogram for girls in
stem,” Computers & Education, vol. 69, pp. 408–418, 2013.
TABLE I. MENTEE’S OPINION OF THE MENTORS’ CHARACTERISTICS [2] J. McCarthy, “International design collaboration and mentoring
for tertiary students through facebook,” ustralasian Journal of
Mentor’s Characteristics Opinion in Percentage
Educational Technology, vol. 28, no. 5, 2012.
Approachable 83%
Accessible 50% [3] E. H. Trainer, A. Kalyanasundaram, and J. D. Herbsleb, “E-
Supportive 83% mentoring for software engineering: a socio technical
Encouraging 83% perspective,” in 2017 IEEE/ACM 39th International Conference
Provides Constructive feedback 42% on Software Engineering: Software Engineering Education and
Provides guidance 42% Training Track (ICSE-SEET). IEEE, 2017, pp. 107–116.
Provides timely response 33% [4] A. Ziegler, “Mentoring: Conceptual basics and effectiveness
Provides comprehensive response 33% analysis,”Mentoring: Theoretical Background, empirical
Suggests appropriate resources 50% Findings and Practical Applications, pp. 7–29, 2009.
[5] KD Mattingly, MC Rice and JL Berge, "Learning analytics as a
The mentees have responded towards agreement of the tool for closing the assessment loop in higher education,"
statement shown as a summary in Tab. I. The mentors have Knowledge management & e-learning: An international journal,
been accessible whenever they needed and recommended vol. 4, no. 3, pp. 236-47, 2012.
them supplementary resources as stated by 50% of the [6] G. Crisp, I. Cruz, “Mentoring college students: A critical review
mentees. To 83% of the mentees the mentors have been easily of the literature between 1990 and 2007”, Research in higher
approachable, supportive and encouraging while only 42% of education, vol. 50, no. 6, pp. 525-45, 2009.
[7] L. Akin and J. Hilbun, “E-Mentoring in Three Voices,” Online
them reported that their mentors provide guidance and
Journal of Distance Learning Administration, vol. 10, no. 1,
constructive feedback. Lastly, 33% have responded that they 2007. [8] E. Gleibermann, A Pedagogy of Mentoring, 2014.
have had timely and comprehensive response from their

88
[8] W.M. Murphy,“From e-mentoring to blended mentoring: education,” Mentoring & Tutoring: Partnership in Learning, vol.
increasing students' developmental initiation and mentors' 24, no. 5, pp. 399–414, 2016.
satisfaction“, Academy of Management Learning & [12] “BigBlueButton Foundation”,Jul.12,2010.Accessed on:
Education, vol.10, no. 4, pp. 606-622,2011. Aug.14,2020.[Online].Available:https://bigbluebutton.org/2010/
[9] I.Roll, N.G. Holmes, J. Day, and D. Bonn. "Evaluating 07/12/bigbluebutton-foundation/
metacognitive scaffolding in guided invention [13] “SLACK Features”. Accessed on:
activities", Instructional science vol. 40, no. 4, pp. 691-710, Aug.14,2020.[Online].Available:https://slack.com/intl/en-
2012.
de/features
[10] Y. Liu, J. Xu and B. Weitz,“The Role of Emotional Expression
[14] Employment and Training Administration United States
and Mentoring in Internship Learning”, Academy of Management
Department of Labor, “Information technology competency
Learning & Education, vol.10, no. 1, pp. 94-110, 2011.
model,” September 2012. [Online]. Available:
[11] E. Argente-Linares, M. C. P´erez-L´opez, and C. Ord´o˜nez-
www.careeronestop.org/competencymodel/info documents/it-
Solana,“Practical experience of blended mentoring in higher
industry.pdf

89
 The Evaluation of Distance Teaching During
COVID Quarantine at MUST
Tsooj Shambaljamts Delgermaa Senden
Uranchimeg Tudevdagva
Department of Mechanical Science Department of Mechanical Science
Faculty of Computer Science
Mongolian University of Science and Mongolian University of Science and
Chemnitz University of Technology
Technology Technology
Chemnitz, Germany
Ulaanbaatar Mongolia Ulaanbaatar Mongolia
[email protected]
[email protected] [email protected]

Abstract—This paper describes the evaluation of distance

teaching during COVID-19 quarantine time at Mongolian II. THE COURSE DECRIPTION
University of Science and Technology, public university,
Mongolia. Due the corona virus almost all universities must A. Traditional way to teach course
be to change teaching type from traditional face to face to According to university statistical information totally
distance, electronic and digital teaching and learning. Many 1264 students studied graduate program of MUST in 2019-
educational institutions have no preparation for such us 2020 academy years. The course "Patent studies" is one of
unexpected change. The Mongolian University of Science
the core course for all graduate students, which has one
and Technology has developed own learning management
credit. This course offers an overview of national and
system for e-learning since 2007. Therefore, e-learning and
distance teaching was not so new for our faculties. But, from
international patent law, and a basic concept of patents,
bachelor to doctor, all level to teach from distance in whole trademark, and copyright. Also, the course more focuses
faculties was not really ready. All faculty members and IT on national patent, trademark, and copyright practice and is
center did what could to fit appeared situation. designed to improve the student's skills like formulate and
draft a patent and trademark application. Throughout the
The "Patent studies" course usually teach in traditional semester each student will assess their knowledge with two
way, in classroom for graduate students. The course has one questionnaire tests.
credit including sixteen hours lectures and eight hours for
self studies. The exam of course is the test exam for 30-40 B. Adaptation of Course for Distance teaching
minutes. All this standard must be adapted to distance
teaching. For the distance teaching we developed sixteen
Due the COVID-19 situation the course "Patent
hours’ video lesson, and spend sixty-four hours in Unimis studies" was deliver from distance. There was not much
LMS, Teams 365 software and FB social networking to time for adaptation of traditional course. However, first of
communicate with our students. all, we converted lecture material into video materials step
by step. After that, all of students have invited into
For evaluation of distance teaching applied the structure Facebook social networking group so that explain and
oriented evaluation model (SURE model). Together with discuss the topics of course. Later on, after discussion and
SURE model we developed structures for evaluation. For course materials, we provide a 15-20 questions self-
data collection we used Google form and collected data assessment test using the QR code. The QR code is
processed by the SURE model online calculator. directly connected to Google form that collects all test
scores in real-time. End of the semester, those test scores
Keywords—evaluation, e-learning, distance teaching, influence the total score. Throughout the semester, the
SURE model. students were divided into groups and each group was
assigned the task of drafting a patent, trademark, and
I. INTRODUCTION copyright application which related to their work and
The Mongolian University of Science and Technology research. Also, each team introduced to a broad range of
(MUST) is the one of basic public university. Most portion patent-related topics via Facebook. End of the semester,
of engineers graduated at MUST. The MUST has 1003 leading presenters lead a discussion for their topics using
teaching staff and 20518 students in 2019-2020 academic Microsoft Teams 365 software.
year. There are 47 faculties offers 294 study directions.
III. THE EVALUATION MODEL
School of Mechanical Engineering and Transportation
is one of the biggest school of the MUST. Today, the There are different evaluation approaches and models
school has been conducting its activity with over 100 are developed for educational evaluation. But, evaluation
faculty members and 2300 students. Department of models which developed for e-learning evaluation is not so
Mechanical Science is the one of main faculty of SoMET, various. The selected evaluation model is the structure
established in 1964, which offered engineering oriented evaluation model for e-learning [5]. This model
fundamental courses for undergraduate and graduate originally developed for evaluation of e-learning process,
students. Since 2016, after the implementation of the therefore fit to our evaluation interest.
CDIO initiative at MUST, professors from the Department
of Mechanical Science have been organized courses more A. The SURE Model
effectively and taught core courses for graduate students Main concept of structure oriented evaluation model
with new teaching methodology and technology such as (SURE model) is based on a multidimensional
blended and a flipped classroom learning [1-4 ].

90
understanding of evaluation process and a corresponding
analysis [6].

B. Steps of Evaluation Process

x Step 1. Definition of key goals. We are defined five
key goals: Teaching materials, Learning
environments, Skills of lecturer, Assessment types
and Acceptance of distance teaching.
x Step 2. Definition of sub goals. To reach key goals
we can define corresponding sub goals. By our
Fig. 3. Questions of sub-goals added to Google form
definition, first key goal defined trough eight sub
goals, second key goals by four sub goals, third key
goals by five sub goals, fourth key goals by six sub x Step 7. Data processing. Collected data processed
goals and fifth key goals by five sub goals by online calculator of the SURE model [8].
respectively. Original data collected by Google form (Step 6). To
use online calculator collected data have to enter
x Step 3. Confirmation of goal structures. Fig.1
into calculator in comma separated vector (CSV)
shows structures of evaluation goals. The structures
form (Fig. 4). After input of data in corresponding
developed in cooperation with SURE expert and
format evaluator should select view of the results.
lecture who teaches course.
There are three group of selection is possible:
o Color scale type. In this group evaluator
can select one of four different versions.
In default mode red-yellow-green
version is selected. That means
evaluation scores will show more in red
color if scores are less than 0.5 and more
green if they are higher than 0.5. This
kind of view is important to stakeholder
to understand and get impression about
evaluation result in short time. Greener
table confirms that evaluation results are
more positive and good. Opposite more
Fig. 1. The structure of evaluation goals. red shows that object evaluated more
worst.
x Step 4. Creation of Checklist. Based on the sub o Checklist data display format. In this
goals adapted checklist is developed (Fig. 2). group evaluator can select one of three
different versions. In default mode non-
normalized version is selected. That
means collected data will appear in
evaluation table in original scale. If
select normalized version, then collected
data shows in evaluation table all from 0
to 1.
o Evaluation table. In this group evaluator
can select one of two different versions.
Fig. 2. The checklist for key goal 1. In default mode empirical evaluation
scores version is selected. That means in
evaluation table will show only final
x Step 5. Confirmation of Checklist. Both side: evaluation scores as empirical values.
lecture of course and SURE expert confirmed
adapted checklist.
x Step 6. Data collection. For data collection used
Google form (Fig. 3).

91
 data processing. Let us do short analyze on calculated
evaluation scores.
x Main result of the evaluation process is 0.97. This is
very high score and by this value we can conclude
that our unplanned distance teaching run almost
excellent. Because highest possible value is 1 and we
have 0.97.
x Next analyze can focus on key goals. How good we
reached our key goals?
B1 = 0.97
B2 = 0.99
B3 = 0.97
B4 = 0.98
B5 = 0.98
Fig. 4. The online calculator of SURE model By evaluation scores we can summarize that all key
x Step 8. Report of evaluation. The result of goals are reached very positive its target. Highest
evaluation process reported to faculties as feedback values received second key goal. Aim of this sub
of unplanned distance teaching period. goal was identify learning environment of distance
learning. Throughout the semester, Unimis Learning
IV. STATISTICS PROCESSING Management System, Teams 365 software, and FB
social networking systems had been used
A. Short Statistics simultaneously to communicate with our students.
The course "Patent studies" offers each semester for x One more deep analyze can do on sub goals
approximately 280-350 freshmen graduate students from evaluation. In total we had 32 sub goals. There are
all of graduate programs of MUST. distributed not linearly by key goals. The fifth key
goal has most many sub goals: 9, that means we did
B. Respoance Data 9 different activities to reach this key goal. In this
Throughout the semester, 68 students were studied in level we can look to detailed result of sub goals.
my group. The checklist was send to all learners. After the Worst evaluation value got sub goal A59. This sub
final exam, approximately 41 students have responded to goal was focused on the Lecturer responsible for
all questions within a week via Google form. distance learning. 30% of total students were
accepted lecturer is not so important the distance
C. Processed Data learning.
Collected data processed by online calculator of SURE
model [7]. Figure 4 shows basic window of online VI. CONCLUSION
calculator. Evaluator should put collected data in CSV The aim of this evaluation was to measure result of
format into the edit field. The online calculator computes distance teaching and learning methodology. The main
all evaluation scores and return results in table (Fig. 5). evaluation score calculated as 0.97. This score shows that
unplanned distance teaching run sacksful in this semester.
All 32 sub goals are evaluated bigger than 70%, means
satisfaction of students are very positive. It gives many
power to faculty to keep and continue distance and online
teaching for our courses.
While the results of the evaluation surveys and the
mean responses of the key goals were obtained from 60%
of total students, it is clear that distance teaching was
significantly influenced students’ ability to learn online
and self-study and digital learning skills. The improvement
confirms the fact that students were adapted online
learning methodology and, thus, learn the self-study by
doing.
In the end we would like to add that to this structure
oriented evaluation checklist we added two open ended
Fig. 5. Part of result screen of online calculator. questions. “What are some ways to increase student
engagement? " and “ What are the first thing you would
like to improve in connection with the lesson?”. We
V. SHORT ANALYZE ON EVALUATION RESULT received interesting different response to these two
questions and actually we are working on reflection and
analyze on those questions. We plan to include analyze and
The advantage of the SURE model is evaluator can comparison of results in our next work.
receive many evaluation scores from evaluation table after

92
 REFERENCES [5] U.Tudevdagva, “Structure Oriented Evaluation Model for E-
Learning”, Wissenschaftliche Schriftenreihe Eingebettete
[1] Delgermaa S. "Opportunity to organize a workshop on “Theoretical Selbstorganisierende Systeme, Universitätsverlag Chemnitz, July
mechanics” using “Case study” technology", ICE 2012. 114 p. 2014. pp. 123. ISBN: 978-3-944640-20-4, ISSN: 2196-3932.
Deajeon, Korea, ''Proceedings of The International Conference on
Engineering'' [6] U.Tudevdagva, & W.Hardt, “A new evaluation model for e-
learning programs”, Technical Report CSR-11-03, Chemnitz
[2] Delgermaa S., Tsogzolmaa D. "Study of Opportunity to Check Out University of Technology.
the Student’s attendance Online", ICKI 2011. 349 p, Ulaanbaatar,
Mongolia, ''Proceedings of International Conference on knowledge [7] U.Tudevdagva, “Structure-Oriented Evaluation An Evaluation
Based Industry'' Approach for Complex Processes and Systems”, Gewerbestrasse
11, 6330 Cham, Switzerland, Springer, 2020, pp.92, ISBN 978-3-
[3] Batbayar D., Delgermaa S., Tsooj. "Evaluating the quality of core 030-44805-9 ISBN 978-3-030-44806-6 (eBook),
courses of Mechanical engineering program" National conference https://doi.org/10.1007/978-3-030-44806-6.
on Outcome based education 2019. pp 105. Ulaanbaatar,
Mongolia [8] Online calculator of the SURE mpdel,
http://uranchimeg.com/sure/eva.php
[4] Ganbat D, Delgermaa S “Some experiences of using ICT in
active learning” ISCSET 2017. Berlin, pp 104-110

93