T4E 2010

A Virtual Laboratory for Distance Education

Kannan M. Moudgalya Inderpreet Arora

Dept of Chemical Engg., IIT Bombay, Mumbai, India Systems and Control, IIT Bombay, Mumbai, India
[email protected] [email protected]

Abstract—With the addition of online laboratories, dis-

tance education can become more meaningful and com-
plete. This paper presents a virtual lab established through
LabVIEW data socket technology that enables the end
user to design and implement control algorithms. Although
this makes the problem more challenging, the quality of
learning is also enhanced enormously. We demonstrate this
lab through an open source, low cost, single board heater
Figure 1. Virtual control lab architecture
system. A large number experiments have been conducted
using this setup.
Keywords-virtual labs; remote access; live data sharing;
they can introduce the device in the new curriculum. Thus,
LabVIEW; real-time control
it makes sense to make available even a low cost device
I. I NTRODUCTION through virtual labs.
The architecture for this lab is explained in Section II.
The fast growing economy demands proactive, knowl- The virtual control lab allows remote access of a single-
edgeable and technically sound professionals and hence board heater system, a multi input single output and open
a major contribution from academia. Distance education, loop stable system, discussed in Section III. Numerous ex-
with the focus on rendering concepts, facilitates “Living periments have been performed on this setup using Scilab
anywhere and studying from anywhere, anytime”. But and LabVIEW. LabVIEW interface for data acquisition
it leaves out practical and application oriented learning and remote access of these experiments is depicted in
components that are required to make a person who has Section II. The conventional virtual labs allow only to
studied theory courses well rounded. change the control parameters remotely with the control
A basic control engineering course covers concepts effort computed at the server end. Section II also explores
and techniques for process modeling and analysis, and the possibility of using virtual labs as a platform for
controller design. It is accompanied by abstract concepts designing and implementing real-time control algorithms
and complicated computations which if supported by where the computation is done at the client end. Such
experiments can hold students’ interest in the course. The strategies need to account for the presence of network
integration of the Internet into education and numerous delay. Section IV describes few experiments performed
web based software packages have strengthened this very on this setup to obtain the model and also to reveal the
idea. Web based control labs have been in educational effect of network delay. The proposed architecture to cater
practice in countries like USA, Singapore, Serbia, China, these needs is depicted in Sectoin V.
Germany and Japan, to name a few, for quite some time
[1], [2]. Under the ongoing Virtual Labs project funded by II. V IRTUAL LAB ARCHITECTURE
MHRD [3], a virtual control lab has been implemented and
tested for teaching undergraduate and postgraduate control A. Hardware architecture
courses, such as Instrumentation and Process Control, and The virtual control lab architecture at IIT Bombay is
Digital Control. shown in figure 1. It resembles the conventional virtual
In this paper, we describe in brief a low cost single- labs that allow remote access of the setup but for live
board heater system that can be put together at a cost of data sharing. The elementary implementation involves the
$50 and which is available also as an open source [4]. We plant, a computer with LabVIEW installation, and an
discuss our experience in using it through virtual labs. internet connection. The client end needs a web browser
We strongly believe that a low cost of a device alone and a LabVIEW runtime engine to remotely access the
does not make it attractive for inclusion in academic setup.
programmes. There is a large inertia to try out any new In order to give a real feel for the system, we explored
device. A remotely operated laboratory could be useful the possibility of hosting an Apache server along with
in this regard. To try out such a system, a prospective the LabVIEW server, for video streaming. Unfortunately,
instructor need not build any infrastructure. They can re- however, it overloads the system, resulting in the loss
motely operate the device and also learn about it from the of video quality. Low bandwidth of 100 kbps required
educational material, also available remotely. If convinced, to cater to students even in remote corners of India [5],

978-1-4244-7361-8/10/$26.00 © 2010 IEEE 190

 Figure 3. Single board heater system

single board heater system, used in this work, is amenable

for this approach, however.
Live data sharing: LabVIEW supports a variety of
Figure 2. Web published front panel of the step response experiment
data sharing methods like UDP/TCP, shared variables,
DataSocket, etc. This eliminates the requirement of con-
introduces a delay of about 7 s. Because of these rea- figuration code in the application. For the above discussed
sons, we abandoned the idea of streaming the video. The virtual control lab, data transfer across computers has been
proposed architecture for the next stage implementation is implemented using DataSocket technology.
likely to take care of audio and video streaming facility NI DataSocket technology allows live data sharing
with advanced and optimized techniques to prevent the with other VIs on local computer or on the web [7].
aforementioned drawbacks. DataSocket is a programming technology based on TCP/IP
communication protocol that facilitates live data transfer
B. Software architecture between different applications on same computer or be-
LabVIEW, a proprietary software by National Instru- tween computers across network. DataSocket comprises
ments, is a graphical programming environment that facili- the DataSocket API and the DataSocket Server which
tates acquisition, storage and display of the measurements, respectively take care of data conversion and streaming.
testing and control of processes, numerical computations, This mode of data sharing requires LabVIEW installa-
etc., using graphical icons and connecting wires. One tion at both the ends. The VI running at the client side
of the salient features of LabVIEW is its networking implements control algorithm and DataSocket server to
capability [6]. LabVIEW VIs (virtual instruments) can communicate with its counterpart at the server end. It takes
communicate with other VIs and processes running on the server ip address as an input argument before starting
other applications or on remote computers. Its web server the communication. The events that take place during this
can host front panel images, embedded VIs and any linked process are explained in PID control implementation in
documentation. With LabVIEW DataSocket Technology, Section IV. The disadvantage of this mode of data sharing
live data can be shared with other VIs across the network. is the requirement of LabVIEW installation at both the
Web publishing tools: LabVIEW web publishing tool ends. This demand does not serve the ultimate aim of
can embed static or animated images of front panel, a VI providing low cost facilities for education and research. A
for remote viewing and control, and also create HTML solution to this problem and a corresponding architecture
documents. The web server needs to be enabled to permit for the single-board heater system virtual lab is explained
client to access these documents. Security features can also in Section V.
be added to allow restricted access. The web published
front panel of the step response experiment is shown in III. S INGLE - BOARD HEATER SYSTEM
figure 2. A single-board heater system, shown in figure 3, is
Most of the existing virtual labs allow remote access of an open source, low cost, lab-in-a-box setup suitable for
the experiment with the entire control lying at the server performing numerous control experiments. It consists of
end only. As a result, the student at the remote end can a heater assembly, fan, temperature sensor, microcon-
only change parameters and observe the resultant profile. troller(ATmega16L) and associated circuitry. This section
Unfortunately, however, the user cannot implement and briefs the hardware and firmware of the apparatus. More
test an algorithm, different from that available at the server details can be sought from [8].
end. To make the learning well rounded, one would have to
give this facility to the student at the remote end so that he Hardware
can try out new control strategies as well. This requires the A 5 cm × 2 cm stainless steel blade whose temperature
control effort to be calculated at the client end, in real time, has to be controlled acts as the plant. Nichrome wire of
depending on the incoming data i.e. after collecting each 0.7 mm diameter wound into a coil of 5 mm × 11 mm
data sample. We have enabled this by using NI DataSocket serves as the heater element. It is positioned at a distance
technology described below. Network delay could pose of 3.5 mm from the steel blade. A low cost computer
problems for difficult systems in the above approach. The fan is meant for cooling the assembly. A monolithic

191
integrated circuit temperature transducer, AD590, soldered
beneath the steel plate produces linear current, 1 µA
per K, proportional to the absolute temperature of the
heated plate. The corresponding voltage reading taken
across 1 kΩ resistor is fed to an instrumentation amplifier.
This circuitry with quad Opamp converts the temperature
reading to ◦ C, raises its strength by a factor of 46 and
feeds it the ADC of ATmega16L. The sensitivity of the
system is improved by operating ADC in 10-bit mode and
restricting the temperature range to 0-90◦ C. The device
communicates with the computer serial or USB port via
UART of ATmega16L. The communicated commands and
input/output data are displayed on LCD. The setup is Figure 4. Local PID implementation
powered by a 12 V DC, 8 A, computer SMPS.

Firmware
The microcontroller, ATmega16L, is the heart of this
setup. It is programmed to convert analog temperature
values to digital, serially communicate commands and
heater, fan inputs and temperature output to and from the
setup with the computer, generate PWM wave to deliver
proportional power to the load and facilitate display of
command and data bytes on LCD. The open source IDE
WinAVR and AVR Studio and the commercial ICCAVR
have been used for programming, compilation and loading
the codes [4].

IV. R EMOTE ACCESS OF CONTROL EXPERIMENTS Figure 5. Client-end panel for PID implementation on internet

Due to complete automation and a small time constant,

we have been able to carry out a large number of ex- PID control implementation
periments using this setup. These include identification
With the live data available through DataSocket tech-
through step, ramp and PRBS testing; Frequency test-
nology, PID control algorithm was implemented locally
ing, leading to Bode plots; PI, PID tuning; 2-DOF pole
and on internet. The motive of this experiment was not
placement, internal model and simple adaptive controllers;
to tune the controller parameters that result in zero steady
Auto tune testing; Self tuning regulator. Data acquisition
state error but to observe the system response for a local
interface has been independently developed using Scilab,
controller and a controller located at a distance.
an open source software, and LabVIEW.
1) Local implementation: In this case, the controller
is located at the acquisition server and the access is also
Step response experiment
done locally. The system response for this implementation
The web published front panel of step test experiment is shown in figure 4. At around 70th sample, heater current
is shown in figure 2. A client needs to install a plugin, a i.e. control effort, increases due to increase in kc, the
LabVIEW 8.0 runtime engine, to enable the browser for controller gain, by the user. The system takes around
remote access and enter relevant experiement url. Once the 16 samples or 5 s to respond. Another change in kc at
connection has been established and panel has been down- around 270th sample causes system to respond in about
loaded, the client can right click and “Request control” 10 samples. The change in setpoint at around 425th sample
from the server. After the control has been granted, the results in 15 samples delayed response.
client is ready to perform the experiment. Same procedure 2) On Internet: In this case, the controller is located at
allows remote access of other available experiments [4]. the client side. The client-end panel for PID implementa-
With fan and heater inputs as manipulated variables and tion is as shown in figure 5. The events occurring at server
temperature as a controlled variable, the client can give and client are discussed next.
a step change in heater current while maintaining fan at The client receives data acquired by the acquisition
a constant speed or vice versa. After the steady state is server through internet with the technology discussed
achieved, another relevant amplitude step change in input earlier. With the PID control algorithm running in the
can be given and a SISO model can be obtained with the client machine and depending on the controller parameters
collected data. Currently, the directory in which the data given by the client, the control effort is calculated. This
are stored, is available only on intranet. However, the data value of control effort, heater current in this case, is sent
are also available, for copy, in the text box on front panel. to the server through internet. Server passes on this control

192
effort to the setup and again acquires temperature from the the setup is low-cost and versatile, the remote access of the
setup, and sends it to the client. Depending on the output setup allows the exploration of the setup before adoption.
temperature, client algorithm again calculates error and As the proposed architecture is generic, it can be extended
the required control effort, and sends it back to the server. to labs that use expensive resources as well.
This goes on continuously. The explanation pertaining to This paper has also discussed the technology for live
plot is similar to that for local controller but for delay by data sharing which in turn leads to real-time control of
number of samples. This delay was found out to be of the experiments. With such platform, researchers can design,
order of 15 s. implement and test algorithms for timer synchronization,
3) Comparison: As expected, network delay plays an network delay prediction and robust control of the exper-
important role in implementation of remotely located con- iments through web.
troller. With local controller causing 5 s delayed response,
ACKNOWLEDGMENT
remote controller was found out to result around 15 s
delayed response. This may not be acceptable for an open The authors would like to thank Mahanand Mali and
loop unstable system or critical applications. A soultion Rupak Rokade for helping with the control experiments,
to this problem is discussed in next section. and Bharat Suthar for setting up Apache server for video
streaming.
V. F UTURE WORK
R EFERENCES
Proposed virtual lab architecture
[1] C. Ko, B. M. Chen, J. Chen, Y. Zhuang, and K. Tan,
The next stage virtual lab architecture along with audio “Development of a web-based laboratory for control exper-
and video streaming will have login facilities for the iments on a coupled tank apparatus,” IEEE Transactions on
users. The user will be required to book the time slot in Education, vol. 44, no. 1, 2001.
advance, login with the authenticated details and perform
[2] M. Stefanovic, V. Cvijetkovic, M. Matijevic, and V. Simic,
the experiments. With one user performing the experiment, “A labview-based remote laboratory experiments for con-
other users, the number depending on the configuration, trol engineering education,” Wiley Periodicals Inc., 2009.
will be allowed to see it. There will also be an open forum
for discussion. For the users having low bandwidth for [3] K. M. Moudgalya, “Introducing National Mission
communication, the audio and video streaming would be on Education through ICT,” http://www.spoken-
tutorial.org/NMEICT-Intro, 2010.
optional thereby taking care of all types of users. For live
data sharing, the possiblity of using LabVIEW executable [4] Virtual labs project, “Single board heater system,”
which when installed at the client end will enable live data http://www.co-learn.in/web sbhs, seen on 15 May 2010.
transfer, is being explored. Use of C, Java, C++, Python,
and GNU Radio for remote access and live data sharing [5] K. M. Moudgalya, R. Deshmukh, and A. Patil, “Syn-
chronous distance education at IIT Bombay,” in Proceed-
are also being explored. ings of the International Workshop on Technology for
Education, T4E 09. Banglore, India: IEEE, Aug. 2009.
Real-time control
The PID control experiments were performed to observe [6] “LabVIEW user manual,” http://www.ni.com/, Tech. Rep.,
the role of network delay while controlling the experi- 2003, National Instruments.
ments through web. Network delay becomes crucial for [7] “Integrating the Internet into your measurement
real-time control applications. The algorithm for real-time system, DataSocket Technical Overview,”
control of experiment through internet can start from timer http://www.ni.com/pdf/wp/wp1680.pdf Seen on 16
synchronization between server and client described in [9]. May 2010, Tech. Rep., National Instruments.
Once the system time at both the ends is synchronized,
[8] I. Arora, K. M. Moudgalya, and S. Malewar, “A low cost,
network delay can be predicted by communicating tokens open source, single-board heater system,” in International
with timestamp [9], [10]. Control strategy can then be Conference on E-Learning in Industrial Electronics. AZ,
devised using time delay compensation method by distur- USA: IEEE, November 2010, in review.
bance observer [11]. The work for timer synchronization
and network delay prediction has commenced. This work [9] S. Wang, B. Xu, Y. Jia, and Y.-H. Liu, “Real-time mobile
robot teleoperation over IP networks based on predictive
will be followed by implementation of robust algorithms control,” in Proceedings of IEEE Int. Conf. on Cont. and
for real-time control. Automation, 2007.
VI. C ONCLUSION [10] T. Shiotsuki, “A new structure for real-time control over
In this paper, a virtual control lab architecture bas been computer network,” in Proceedings of Int. Conf. on
Mechantronics, 2007.
presented. The architecture being simple and cost effec-
tive, has the potential to bridge the gap between theory and [11] K. Natori and K. Ohnishi, “An Approach to Design Feed-
practice and improve the quality of conventional distance back Systems with Time Delay,” in Proceedings of the
education in the area of control. The paper also has briefly 31st Annual Conference of the IEEE Industrial Electronics
presented a single-board heater system, a low-cost control Society, North Carolina, USA, Nov. 2005.
setup used for performing experiments remotely. Though

193