Hindawi

Wireless Communications and Mobile Computing

Volume 2022, Article ID 8274455, 13 pages
https://doi.org/10.1155/2022/8274455

Research Article
Robot Indoor Positioning and Navigation Based on Improved
WiFi Location Fingerprint Positioning Algorithm

Hemin Ye and Jiansheng Peng

Hechi University, 42 Longjiang Road, Yizhou 546300, China

Correspondence should be addressed to Jiansheng Peng; [email protected]

Received 16 August 2021; Revised 14 December 2021; Accepted 5 January 2022; Published 31 January 2022

Academic Editor: Pietro Manzoni

Copyright © 2022 Hemin Ye and Jiansheng Peng. This is an open access article distributed under the Creative Commons
Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work
is properly cited.

In response to the traditional WiFi location fingerprint positioning algorithm still having a low positioning accuracy, which is
difficult to meet the robot indoor positioning and navigation needs, a series of improvements are made to the traditional WiFi
location fingerprint positioning algorithm, so that the positioning accuracy of the algorithm can be effectively improved. At the
stage of building the location fingerprint library offline, WiFi signals are collected at each reference point by reducing the
reference point spacing of the traditional location fingerprinting algorithm and then using different time period collection
methods. The WiFi signal strength values are standardized using the standardization processing method to improve the
specificity of traditional location fingerprinting. In the real-time localization stage, the WiFi signals collected from the
unknown location points are averaged, and then, the fingerprint similarity calculation is performed using a matching method
based on the magnitude of the Marxian distance as a similarity reference. In order to eliminate the location fingerprints that
degrade the localization accuracy, an improved adaptive K-value WKNN algorithm is integrated at the end of the localization
algorithm. The improved localization algorithm and the proposed raster-based navigation algorithm are validated in a fixed
experimental environment. The experimental results show that the probability of the improved algorithm’s positioning error
within 0.4 m is 49%, which is a 35% improvement over the conventional algorithm. Combining the improved positioning
algorithm with our proposed grid-based navigation algorithm, the final navigation error probability within 0.8 m is 62%.

1. Introduction and accelerometer to develop a trackless food delivery robot

with navigation function. The application scenarios of the
In recent years, due to the development of sensors, chips, and factory handling robots, hospital medical robots, hotel ser-
artificial intelligence, robots have become more and more vice robots, and restaurant delivery robots introduced above
intelligent and flexible [1]. Therefore, robots are used in more are shown in Figure 1. For these robots to successfully
and more fields. Hata et al. developed a robot system with complete these specific tasks, the robot’s positioning and
flexible manipulators, stereo cameras, and 3D vision for cloth navigation functions are indispensable. The robot with the
handling [2]. A company based in Pittsburgh, Philadelphia, positioning and navigation function can perceive its current
Aethon, has developed a medical mobile robot that can help position and can navigate to the specified position in the
patients in the ward transport needed food and cleaning sup- above application scenarios.
plies, and it can also accurately and timely provide prescrip- The popular GPS (global positioning system) [6] can
tion drugs and medical supplies to nursing stations [3]. Now, obtain high positioning accuracy outdoors. The civilian use
many robots are also used in hotels. These robots can clean, can reach a positioning accuracy of 3 meters, and the mili-
move items, provide information, accept customer orders, tary use can even achieve a positioning accuracy of 0.3
process payments, etc. [4]. Literature [5] proposes multi- meters. GPS is widely used in the positioning and navigation
source information fusion, which combines ultrawideband of moving objects such as automobiles, ships, airplanes, and
positioning technology, an odometer, low-cost gyroscope, handheld mobile terminals. However, most of the robots in
 6302, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2022/8274455 by Readcube (Labtiva Inc.), Wiley Online Library on [08/10/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2 Wireless Communications and Mobile Computing

Factory handling robot Hospital medical robot Hotel service robot Restaurant delivery robot

Figure 1: Robots in various application scenarios.

the above application scenarios are in complex indoor envi- 2. Related Work
ronments, and it is almost impossible to apply GPS to
indoor positioning. The biggest problem is that the GPS sig- Robot indoor positioning technology first appeared in the
nal will be greatly weakened when it encounters buildings middle and late 20th century. At that time, researchers from
and various indoor objects, so the positioning error will Japan proposed a reliable, safe, and cheap positioning system
increase sharply when GPS is used for indoor positioning for indoor mobile robots [8]. It uses an active ultrasonic bea-
[7]. Therefore, it is necessary to find a positioning technol- con installed on the ceiling to send an ID code to the robot
ogy that is suitable for indoor positioning and has high posi- via a wireless link to control the transmission of ultrasonic
tioning accuracy and low cost. It is of important research pulses. The two receivers on the robot measure the flight
significance to apply this positioning technology to the time of ultrasonic pulses from two beacons at known posi-
robot’s indoor positioning so as to realize the robot’s indoor tions. After getting the flight time, the two-dimensional
positioning and navigation. position and direction of the robot can be determined. The
With the development of wireless local area networks, developed system enables the robot to obtain good position-
WiFi devices have become network devices that can be seen ing accuracy in a small area. However, the ultrasonic waves
everywhere indoors. Using indoor WiFi signals to locate used in the system are easily affected by indoor complex fac-
robots has the advantages of low cost, high accuracy, large tors, which may lead to increased positioning errors, and the
positioning coverage, high indoor penetration, and high ultrasonic transmission distance is short. Therefore, it can
communication capabilities. However, the current tradi- only work within a certain range. This limits the develop-
tional WiFi location fingerprint positioning algorithm is ment of ultrasonic positioning technology.
generally not high in positioning accuracy, and it is difficult In order to solve this problem, researchers try to use
to meet the requirements of robots for indoor positioning lidar to obtain information about the indoor environment,
accuracy. For this reason, this article makes corresponding thereby realizing the positioning of the robot. In the litera-
improvements on the basis of the traditional WiFi location ture [9], a robot uses the raw data provided by two laser ran-
fingerprint positioning algorithm, so that the positioning gefinders to estimate the distance and angle to another
accuracy of the improved positioning algorithm can meet robot, thereby realizing multirobot positioning. However,
the requirements of the robot for indoor positioning accu- using lidar alone for robot positioning, the positioning accu-
racy. In order to enable the robot to navigate to the desig- racy is not particularly ideal. In order to further improve the
nated position smoothly, the author proposes a grid-based positioning accuracy of the robot, people try to integrate the
navigation algorithm based on the characteristics of the lidar with other sensors for positioning.
improved positioning algorithm. Compared with most nav- Researchers from the Beijing Institute of Technology
igation algorithms, this navigation algorithm has the advan- proposed a positioning and navigation system based on the
tages of simple implementation, less calculation, and higher fusion of lidar and camera [10]. They proposed a three-
accuracy. dimensional (3D) semantic map with large-scale and accu-
The rest of the paper is organized as follows. The second rate integration of lidar and camera information to realize
part is a literature review of the robot’s indoor positioning real-time road scenes. Moreover, a convolutional neural net-
and navigation. The third part introduces the system design, work- (CNN-) based semantic segmentation of images is
including the design of the robot, the design of the position- employed to develop the semantic map of the environment.
ing model, and the design of the navigation model. The On the basis of time and space synchronization, the sensor
fourth part discusses the improved WiFi location fingerprint fusion of lidar and camera is used to generate semantic mark
positioning algorithm and the grid navigation algorithm frames of the point cloud and then generate a semantic map
based on the positioning algorithm. The fifth part presents based on attitude. Literature [11] presents a mobile robot
our experimental design and results. The sixth part is the simultaneous localization and mapping method that com-
conclusion of the full text. bines lidar and inertial measurement unit to set up a
 6302, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2022/8274455 by Readcube (Labtiva Inc.), Wiley Online Library on [08/10/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wireless Communications and Mobile Computing 3

multisensor integrated system and uses a rank Kalman filter- Detect

ing to estimate the robot motion trajectory through inertial Motor Encoder
measurement unit and lidar observations. Literature [12]
PWM Read
utilizes data gathered with an autonomous delivery robot control pulse
to compare different sensor fusion techniques and evaluate
PID algorithm
which algorithms provide the highest accuracy depending
on the environment. They show how lidar scan matching
Raspberry Serial port STM32f103RCT6
combined with other sensor data can be used to increase pi
the accuracy of the robot localization and, in consequence, (OpenWrt) Obstacle Grid
its navigation. Although the fusion positioning by lidar and avoidance navigation
other sensors can obtain a higher positioning accuracy, due algorithm algorithm
to the high price of lidar, it is difficult for this positioning
technology to be popularized and developed. Detect
In recent years, with the development of artificial intelli-
gence and machine vision, more and more researchers Infrared obstacle
avoidance module
choose to use visual SLAM- (simultaneous localization and
mapping-) based [13] robot indoor positioning and naviga-
tion technology. For example, in literature [14], robots are Figure 2: System model diagram.
applied to hospitals, and a SLAM technology based on
RGB and depth images is proposed to improve hospital
operation efficiency. Since most of the current visual SLAM start the raster navigation algorithm for navigation. During
researches assume that the environment is static, it will be a the navigation process, the robot’s walking speed needs to
challenge when dealing with real and dynamic scenes. Based be controlled and maintained at a constant speed value in
on this problem, literature [14] also proposed a method that order to reach the actual destination more accurately. There-
effectively deals with SLAM problems for scenarios with fore, the pulses generated by the motor rotation are detected
dynamic objects, e.g., people and movable objects, based using an encoder. Then, a PWM [19] control is finally
on the semantic descriptor extracted from images with help applied to the motor through the calculation of a PID algo-
of a knowledge graph. Literature [15] studied real-time robot rithm [20]. This is the closed-loop control of the PID algo-
detection and map construction. The autonomous robot rithm. The navigation process due to the positioning error
uses a single stereo vision camera to navigate, self-position, at the starting point may cause the robot to walk away from
define the environment, and avoid any possible obstacles the intended route and thus encounter obstacles, which
to follow the best route to the greatest extent to draw the requires the obstacle avoidance algorithm for obstacle avoid-
map area. Recently, the research of visual SLAM positioning ance. The obstacle avoidance algorithm designed in this
and navigation in the field of robotics remains high due to its paper uses an infrared obstacle avoidance module for obsta-
own advantages. However, its shortcomings are also more cle detection.
prominent, such as the following: visual SLAM is more sen-
sitive to light, and it is easy to fail to match when the light is 4. Methodology
not good; the image processing of visual SLAM is compli-
cated; the amount of calculation is large; and the processing Figure 3 shows the model diagram of the improved WiFi
speed is slow. For this reason, this article combines the location fingerprinting algorithm. The algorithm is divided
improved WiFi location fingerprint positioning algorithm into an offline phase to build the location fingerprint data-
and the proposed grid-based navigation algorithm for posi- base and a real-time localization phase. In the phase of
tioning navigation, which may be a new mode to solve the building the location fingerprint database offline, the indoor
above difficulties, because the improved positioning algo- is first divided into equally spaced raster points (see 4.1.1),
rithm in this paper has achieved good results in positioning and the spacing of each raster point is smaller than the raster
accuracy. In addition, the proposed grid navigation algo- point spacing of the traditional algorithm. The robot per-
rithm is simple to implement, small in calculation, fast in forms WiFi signal acquisition at each grid point (reference
processing speed, and not affected by the above-mentioned point) in the room, and unlike the traditional algorithm,
factors. the improved algorithm uses WiFi signal acquisition
methods for different time periods (see 4.1.2). The tradi-
3. System Design tional algorithm is prone to matching errors due to the low
specificity of the location fingerprint. Therefore, the col-
The designed positioning algorithm is running on the lected data are normalized [21] (see 4.1.4) to improve the
OpenWrt [16] system on board the Raspberry Pi [17]. How- location fingerprint specificity and thus the matching accu-
ever, it is the STM32 [18] that controls the robot’s motion. racy. In the real-time localization phase, the robot acquires
Therefore, the Raspberry Pi transmits the current position- a real-time WiFi signal at an unknown point in the room.
ing result and destination point information to the STM32 The collected WiFi signals are then averaged. The formula
by means of serial communication, as shown in Figure 2. of the Mahalanobis distance [22] is used in the matching
When the STM32 gets the data from the serial port, it will algorithm to calculate the Mahalanobis distance between
 6302, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2022/8274455 by Readcube (Labtiva Inc.), Wiley Online Library on [08/10/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
4 Wireless Communications and Mobile Computing

Divide the Standardized

equally spaced Collect WiFi processing of WiFi
grid signal signals

Location
Reference Offline establishment of fingerprint
point(RP) location fingerprint database
database stage

Get real-time WiFi Mean Matching

signal processing algorithm

Improved adaptive
Real-time K value
Reference positioning WKNN algorithm
point(RP) stage

Position
coordinates

Figure 3: Improved WiFi location fingerprint positioning algorithm model diagram.

the data collected from this unknown point and each loca- 4.1.2. WiFi Signal Collection Method. WiFi signals are
tion fingerprint in the location fingerprint database. The size affected by indoor obstructions, signal multipath propaga-
of the Mahalanobis distance is used as a reference for the tion, environmental factors, and other conditions. Therefore,
similarity, and the smaller the Mahalanobis distance, the WiFi signal strength values fluctuate from time to time, and
greater the similarity. Finally the improved adaptive K the degree of fluctuation varies at each time of the day. The
-value WKNN [23] algorithm (see 4.2.3) is fused to remove traditional location fingerprinting algorithm collects WiFi
the noisy points, thus improving the accuracy of location signals at one time of the day and does not take into account
estimation. When the estimated coordinates of the unknown the fluctuation of WiFi signal strength values at different
point are obtained, they are sent to the mobile or PC termi- times of the day, which will have a significant impact on
nal. The user can view the robot’s position in the room on the positioning accuracy if these data are used to build the
the control platform, indicated by a red dot. fingerprint database. To address this problem, the improved
algorithm collects 100 WiFi signals at each reference point in
the indoor environment to build a location fingerprint data-
4.1. Offline Establishment of Location Fingerprint base. However, these 100 collections were made at different
Database Stage times of the day, and the number of collections for each time
4.1.1. Divide Equidistant Grid. Traditional location finger- period is shown in Table 1. Such a collection method will
printing algorithms usually perform equidistant raster divi- effectively solve the problem of large fluctuations of WiFi
sion of the indoor environment before collecting WiFi signals on the localization accuracy compared with the col-
signal data offline, but the reference point interval divided lections method of traditional algorithms.
is usually 1 meter and above. Experimentally, we found that
the localization accuracy of the same algorithm is improved 4.1.3. Determination of the Number of WiFi Devices. In order
as the reference point spacing gradually decreases. Consider- to find out the number of WiFi devices with the best local-
ing the algorithm localization accuracy and data collection ization accuracy performance in the experimental environ-
workload, we decided to use a 0.3 m equidistant raster in ment of this paper, six groups of experiments were done in
the improved WiFi location fingerprinting algorithm. the indoor environment of Figure 4. First, one WiFi device
Figure 4 shows the top view of the room with a reference is selected to build the fingerprint database and real-time
point spacing of 0.3 m. All subsequent experiments will be localization, and then, the number of WiFi devices is gradu-
conducted in this environment. ally increased. The WiFi devices are added in the order of
 6302, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2022/8274455 by Readcube (Labtiva Inc.), Wiley Online Library on [08/10/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wireless Communications and Mobile Computing 5

AP1

0.3m
AP5

AP4

AP6
AP3

AP2

Figure 4: Distribution map of reference points with a 0.3-meter spacing.

Table 1: The number of collections corresponding to each period.

point t after the normalization process, and t represents the
current reference point (t = 1, 2, ⋯, N). For such fingerprint
Time period Collection times data, there are N in the location fingerprint database (N rep-
8:00-11:30 in the morning 30 resents the total number of reference points). Among them,
ðxt , yt Þ represents the coordinates of the reference point t
12:00-1:30 noon 10
room. mðtÞ is the expected vector of 4 WiFi signal strengths
2:30-5:30 in the afternoon 30
collected at the reference point t. mðtÞ can be estimated from
7:30-10:30 in the evening 30
the following:
!
WiFi device numbers. Figure 5 shows the average localiza- n n n n
1 1 1 1
tion error of the improved localization algorithm with differ- mð t Þ = 〠 Xt , 〠 Xt , 〠 Xt , 〠 Xt : ð1Þ
n k=1 1,k n k=1 2,k n k=1 3,k n k=1 4,k
ent numbers of WiFi devices. It can be seen from the figure
that the average positioning error of the algorithm does not
decrease significantly when the number of WiFi devices is The first item in the vector represents the average value
increased from 4 to 5 or even 6. Based on this conclusion, of the signal strength of the access point AP1 collected at
4 WiFi devices are used in the subsequent experiments to the reference point t, X represents the signal strength value,
ensure a high positioning accuracy with a minimum number k represents the kth acquisition, and n is the total number of
of WiFi devices. acquisitions. The second, third, and fourth items, respec-
tively, represent the average signal strength of AP2, AP3,
4.1.4. Standardized Processing of WiFi Signals. The location and AP4. ∑t is a covariance matrix, which can be expressed
fingerprint data constructed in the offline phase of the tradi- as
tional location fingerprint positioning algorithm is expressed 0 1
as ðxt , yt Þ, rssi1 , rssi2 , rssi3 , and rssi4 , which contains the σ2t,1 λt1,2 λt1,3 λt1,4
coordinates of the reference point t and the average value B t C
t    B λ2,1 σ2t,2 λt2,3 λt2,4 C
of the collected 4 WiFi signal strengths. However, the speci- 〠 = cov X ti , X tj B C:
= B C
ficity of the location fingerprint constructed by the tradi- i, j=1,4 B λt λt3,2 σ2t ,3 λt3,4 C
@ 3,1 A
tional location fingerprinting algorithm is not high, so it is
easy to generate matching errors during localization, which λt4,1 λt4,2 λt4,3 σ2t ,4
reduces the algorithm localization accuracy. In order to ð2Þ
improve the specificity of the location fingerprint and thus
the accuracy of the algorithm, a standardized processing Among them, λti, j ≡ cov ðX ti , X tj Þ = EðX ti , X tj Þ − EðX ti Þ ×
method is used to standardize the collected WiFi signals
EðX tj Þ, i, j = 1,4ði ≠ jÞ; λti,i ≡ σ2t ,i = VarðX ti Þ = EððX ti Þ Þ −
2
and thus build a standardized location fingerprint database.
ððxt , yt Þ, mðtÞ , ∑t Þ is the location fingerprint of the reference ðEðX ti ÞÞ , i = 1,4.
2
 6302, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2022/8274455 by Readcube (Labtiva Inc.), Wiley Online Library on [08/10/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
6 Wireless Communications and Mobile Computing

1.0

Average positioning error (m) 0.8

0.6

0.4

0.2

0.0
1 2 3 4 5 6
Number of APs

Figure 5: A schematic diagram of the influence of different AP numbers on the positioning accuracy of the algorithm in the experimental
environment of this article.

The estimated value of ∑t at the reference point t is λti, j = iAP2 , rssiAP3 , rssiAP4 Þ of the unknown position can be
cov ðX ti , X tj Þ, i, j = 1,4, and the estimated value of λti, j can obtained.
be expressed as follows:
rssi1APn + rssi2APn +⋯+rssikAPn
rssiAPn = : ð4Þ
! ! k
1 n t  1 n t 1 n t
αtij = 〠 X i,k × X tj,k − 〠X × 〠X : ð3Þ 4.2.2. Matching Method. After obtaining the signal
n k=1 n k=1 i,k n k=1 j,k strength vector x = ðrssiAP1 , rssiAP2 , rssiAP3 , rssiAP4 Þ of the
unknown location point, the Mahalanobis distance is used
as the standard to measure the similarity between x and
∑t is estimated as ∑t = ðαtij Þi, j=1,4 . The covariance matrix the location fingerprint data in the matching method.
∑t is a symmetric matrix ðαtij = αtji Þi, j=1,4 . So just calculate The smaller the Mahalanobis distance, the greater the sim-
ilarity between the corresponding location fingerprint and
the upper or lower triangle of the covariance matrix. x. On the contrary, the larger the Mahalanobis distance,
During real-time positioning, the robot collects real-time the smaller the similarity between the corresponding loca-
WiFi signal strength values at unknown coordinate points. tion fingerprint and x. The Mahalanobis distance formula
Then, the coordinates of the unknown location are obtained can be expressed as follows:
by calculating the currently collected signal strength value
and mðtÞ and ∑t in all fingerprint data in the location finger- vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
!−1
u
uÀ Á t
À Á
print database. The specific process will be introduced in the
d= t x−m × 〠 ð t Þ × x − mðt Þ T , ð5Þ
real-time positioning phase.

4.2. Real-Time Positioning Stage where x is the real-time WiFi signal strength vector col-
lected by the robot at an unknown location, mðtÞ is the
4.2.1. Signal Collection and Processing. In the real-time posi- expected WiFi signal strength vector of the location finger-
tioning stage, the robot first collects WiFi signals 5 times at print of the reference point t, and ∑t is the covariance
unknown locations and then performs average processing matrix of the location fingerprint of the reference point t.
according to formula (4), where rssiAPn represents the
average WiFi signal strength of the nth (n = 1, 2, ⋯, 4) 4.2.3. Improved Adaptive K-Value WKNN Algorithm. The
AP; rssikAPn represents the WiFi signal strength value of position fingerprint of the minimum Marxian distance can
the nth AP collected at the kth (k = 1, 2, ⋯, 5) time. After be calculated by Equation (5). However, using the coordi-
such processing, the signal strength vector x = ðrssiAP1 , rss nates corresponding to the minimum Marxian distance
 6302, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2022/8274455 by Readcube (Labtiva Inc.), Wiley Online Library on [08/10/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wireless Communications and Mobile Computing 7

Figure 6: Navigation page.

location fingerprint as the estimated coordinates of the ing positioning accuracy, these K location fingerprints are
unknown location point will make the error fluctuation of eliminated for the second time. First, we calculate the
the algorithm localization larger, thus increasing the local- Euclidean distance between the coordinates ðx1 , y1 Þ corre-
ization error. To solve this problem, an improved adaptive sponding to the fingerprint of the smallest Mahalanobis dis-
K-value WKNN algorithm is fused in the final stage of the tance and the fingerprint coordinates of the other K − 1
algorithm. First, we use Equation (5) to calculate the Maha- locations. They are marked as dist12 , dist13 ,…, dist1k , where
lanobis distance d between x and all location fingerprints. dist1k can be expressed as
Then all Mahalanobis distances d are sorted in ascending qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
order. After sorting, the smallest Mahalanobis distance is
dist1k = ð x 1 − x k Þ2 + ð y 1 − y k Þ2 : ð8Þ
d 1 , the largest Mahalanobis distance is dN , N is the number
of location fingerprints, and d1N is the difference in Mahala-
nobis distance between d 1 and dN . We calculate the average When dist1k > dmin , the location fingerprint will be
value of Mahalanobis distance difference from the following removed from the K location fingerprints. The value of
equation: d min can be adjusted according to different environments.
After testing, better results can be obtained in the experi-
  mental environment of this article when dmin = 1.
d 12 + d13 +⋯+d 1ðN−1Þ + d 1N Finally, the coordinates corresponding to the remaining
d = : ð6Þ k (k < K) position fingerprints are weighted:
N −1

The standard deviation S of the Mahalanobis distance k

1/ðd t + εÞ
difference is calculated by the following: ðx, yÞ = 〠 k
ðx t , y t Þ , ð9Þ
t=1 ∑t=1 ð1/ðd t + εÞÞ
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
À Á2 where ðx, yÞ is the final positioning result obtained by the
∑Nt=2 d 1t − d
S= : ð7Þ robot, ðxt , yt Þ is the coordinate of the reference point t, dt
N −1
is the Mahalanobis distance corresponding to the finger-
print at the reference point t, and ε is a small positive inte-
If d1t > S, then the corresponding location fingerprint ger, and its function is to prevent the denominator from
will be excluded. Otherwise, keep it, and the number of loca- being zero.
tion fingerprints at the remaining is K.
In order to further eliminate outlier location fingerprints 4.3. Navigation Method. In the stage of constructing the
and retain location fingerprints that are effective for improv- location fingerprint database of the positioning algorithm,
 6302, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2022/8274455 by Readcube (Labtiva Inc.), Wiley Online Library on [08/10/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
8 Wireless Communications and Mobile Computing

the indoor environment is divided into equidistant grids.

Start
The vertices of each grid are used as reference points when
building the location fingerprint database, and each refer-
ence point has a unique coordinate. The vertex at the upper
left corner of the indoor plan is used as the origin of the
Whether to receive a
coordinate axis. The right direction of the origin is the pos- navigation request
itive direction of the x-axis, and the bottom of the origin is N
the positive direction of the y-axis. In this design, the refer- Y
ence point spacing is set to 0.3 meters.
In the navigation process, a grid navigation algorithm is Location
used. First, the program has been waiting for the user’s nav- algorithm
igation request. When the user passes through the naviga-
tion page of the robot control platform as shown in
Figure 6, the navigation destination is entered into the web Difference between
page. After the user clicks to start navigation, the Raspberry horizontal and vertical
coordinates
Pi OpenWrt system can receive the user’s navigation request
at this time. Then, the robot starts the positioning program
to obtain the coordinates of the current position, and the
Calculate the
coordinates of the destination point and the coordinates of displacement time in
the current position are made different. The displacement the x and y
duration of the robot in the x direction and y direction is cal- directions
culated according to the difference result and the walking
speed of the robot, as shown in Equations (10) and (11).
Start
X − Xs navigating
Tx = e , ð10Þ
v
Ye − Ys
Ty = , ð11Þ Whether to encounter N
v
obstacles
where X e is the abscissa of the destination point, X s is the
abscissa of the current location point, v is the robot walking Y
speed, T x is the displacement control time in the x direction,
Y e is the ordinate of the destination point, Y s is the ordinate Obstacle
of the current location point, and T y is the displacement in avoidance
end
the y direction control time. Then, according to the displace- algorithm
ment time in each direction, the robot is controlled to move
from the starting point to the target point. During the move-
Figure 7: Grid navigation method program flow chart.
ment, the robot always turns on the function of detecting
obstacles. When obstacles are encountered during naviga-
tion, the obstacle avoidance algorithm is activated. After
fingerprint positioning algorithm is usually 1 meter, 1.5
solving the obstacle problem, the positioning program will
meters, etc. In order to verify the effectiveness of reducing
be restarted to obtain the current position coordinates, and
the reference point spacing to improve the positioning
then, the process just now continues to loop. If there are
accuracy of the improved WiFi location fingerprint posi-
no obstacles during navigation, the robot can reach the
tioning algorithm, we compared the positioning perfor-
end point smoothly and end the program. Figure 7 is a pro-
mance of the improved WiFi location fingerprint
gram flow chart of the grid navigation method.
positioning algorithm when the reference point spacing is
1.5 meters, 1 meter, 0.5 meters, and 0.3 meters. Figure 9
5. Experimental Design and Results shows the comparison of the cumulative error distribution
In order to verify the effectiveness of the improved WiFi functions of the four cases. It can be seen from the figure
location fingerprint location algorithm, we conducted a that as the distance between the reference points gradually
series of experiments in Laboratory 508, Comprehensive decreases, the positioning performance of the improved
Building of Hechi University. Figure 8 shows the real exper- WiFi location fingerprint positioning algorithm is getting
imental environment of Figure 4. The laboratory is 10.5 better and better. Compared with the four reference point
meters long and 7.3 meters wide. In this indoor environ- spacing situations, when the reference point spacing is 0.3
ment, 4 WiFi devices are arranged in different locations. meters, the algorithm in this paper can achieve a better
positioning effect. Theoretically speaking, reducing the dis-
5.1. Experiments with Different Reference Point Spacing. The tance between the reference points again may result in a
reference point spacing set by the traditional WiFi location better positioning effect. However, considering that the
 6302, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2022/8274455 by Readcube (Labtiva Inc.), Wiley Online Library on [08/10/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wireless Communications and Mobile Computing 9

AP1
Robot
AP3 AP4

AP2

Robot

Figure 8: Real indoor environment.

1.0

0.8
Cumulative distribution function

0.6

0.4

0.2

0.0
0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2 3.6 4.0 4.4 4.8
Error (m)
1.5 m reference point spacing 0.5 m reference point spacing
1 m reference point spacing 0.3 m reference point spacing

Figure 9: Comparison of the cumulative error function of the algorithm in this paper in the case of 4 kinds of reference point spacing.
 6302, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2022/8274455 by Readcube (Labtiva Inc.), Wiley Online Library on [08/10/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
10 Wireless Communications and Mobile Computing

−30

−35

WiFi signal strength value

−40

−45 −44

−50

−55

−57
−60
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
Collection times

AP_1 AP_3
AP_2 AP_4

Figure 10: The signal strength values of 4 WiFi devices collected by the robot at the same reference point.

1.0

0.8
Cumulative distribution function

0.6

0.4

0.2

0.0
0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2 3.6 4.0 4.4 4.8
Error (m)
Collect data at a fixed period of time
Collect data at different times

Figure 11: Comparison of positioning effect between different time period collection methods and fixed time period collection methods.

smaller the reference point, the greater the workload of col- 5.2. Collect Data at Different Time Periods and Collect Data
lecting data, therefore, after comprehensive consideration, at Fixed Time Periods. Figure 10 shows 100 data of 4 WiFi
0.3 meters is used as the reference point distance of the devices collected by the robot at a reference point. These
algorithm in this paper. 100 data are collected in different time periods. It can be
 6302, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2022/8274455 by Readcube (Labtiva Inc.), Wiley Online Library on [08/10/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wireless Communications and Mobile Computing 11

1.0

0.8

Cumulative distribution function

0.6

0.4

0.2

0.0
0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2 3.6 4.0 4.4 4.8
Error (m)
Traditional WiFi location fingerprint positioning algorithm
Improved WiFi location fingerprint positioning algorithm

Figure 12: Comparison of the cumulative error function of the two algorithms.

1.0

0.8
Cumulative distribution function

0.6

0.4

0.2

0.0
0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2 3.6 4.0 4.4 4.8
Error (m)
Robot navigation
Improved WiFi location fingerprint positioning algorithm

Figure 13: Cumulative distribution function of navigation error.

seen from the figure that the WiFi signals of the four WiFi time, and the number of times to collect data is relatively
devices fluctuate greatly at different times of the day, and small. Therefore, the collected data is not complete. If the
the maximum fluctuation range of the WiFi signal of AP4 time of real-time positioning is different from the time
is -44~-57. However, the traditional WiFi location finger- period of data collection, it will bring greater positioning
print algorithm usually collects data in a fixed period of error.
 6302, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2022/8274455 by Readcube (Labtiva Inc.), Wiley Online Library on [08/10/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
12 Wireless Communications and Mobile Computing

In order to verify the positioning effect of data collected positioning algorithm, so that its positioning accuracy can
at different time periods and data collected at fixed time meet the accuracy requirements of the robot’s indoor posi-
periods, we adopted different time period collection methods tioning and navigation. In order to improve and optimize
and fixed time period collection methods in the improved the traditional WiFi fingerprint positioning algorithm, we
WiFi location fingerprint positioning algorithm. The final research and analyze which factors will affect the positioning
cumulative error distribution function of the two collection accuracy of the traditional WiFi fingerprint positioning algo-
methods is shown in Figure 11. It can be clearly seen from rithm, and we have verified the influence of these factors on
the figure that the positioning effect of the different time the positioning accuracy of the traditional WiFi fingerprint
period collection method is better than the fixed time period positioning algorithm through experiments. Based on these
collection method. influencing factors, the traditional WiFi location fingerprint
positioning algorithm has been improved as follows: the ref-
5.3. Comparison of Improved WiFi Location Fingerprint erence point spacing of the traditional algorithm is reduced,
Positioning Algorithm and Traditional WiFi Location the WiFi signal is collected by the different time period col-
Fingerprint Positioning Algorithm. The improved WiFi loca- lection method, the standardized processing method is used
tion fingerprint positioning algorithm in this paper has to construct the standardized location fingerprint database,
made a series of improvements on the basis of the traditional the matching method using the Mahalanobis distance as
WiFi location fingerprint positioning algorithm. In order to the similarity reference is adopted, and the improved adap-
verify the positioning effect of the improved algorithm, we tive K-value WKNN algorithm is integrated. In addition,
carried out the experiment of the improved WiFi location based on the improved WiFi location fingerprint positioning
fingerprint location algorithm and the experiment of the tra- algorithm, we propose a grid navigation algorithm to achieve
ditional WiFi location fingerprint location algorithm under robot indoor navigation. The final experimental results show
the same experimental environment. The cumulative error that the probability of positioning error within 0.4 m and
distribution functions of the two algorithms are shown in 0.2 m of the improved positioning algorithm is 49% and
Figure 12. It can be seen from the figure that the probability 39%, respectively, which is 35% and 39% higher than that
of the traditional WiFi location fingerprint positioning algo- of the unmodified positioning algorithm. In addition, the
rithm positioning error below 0.4 meters is 14%, and the probability of indoor navigation error within 0.8 meters
improved WiFi location fingerprint positioning algorithm can reach 62%. These experimental results are sufficient to
positioning error probability is 49% below 0.4 meters. The show that the robot indoor positioning and navigation based
probability of positioning error within 0.4 meters is on the improved WiFi location fingerprint positioning algo-
increased by 35% on the basis of traditional algorithms. In rithm is feasible.
addition, the probability of the traditional WiFi location fin-
gerprint positioning algorithm with a positioning error Data Availability
within 0.2 meters is 0, and the improved WiFi location fin-
gerprint positioning algorithm with a positioning error The data used to support the findings of this study are
within 0.2 meters has a probability of 39%. The probability included within the article.
of positioning error within 0.2 meters has increased by 39%.
Conflicts of Interest
5.4. Robot Navigation Experiment. In order to verify the fea-
sibility of the improved WiFi location fingerprinting algo- The authors declare that they have no conflicts of interest to
rithm for indoor positioning and navigation of the robot, report regarding the present study.
we conducted several experiments in the absence of obsta-
cles and with obstacles. It was found that the actual arrival Acknowledgments
position of the robot has a certain deviation from the given
The authors are highly thankful to the National Natural
end position, which is caused by a certain error in position-
Science Foundation of China (No. 62063006), the Natural
ing and the accumulation of errors during the robot naviga-
Science Foundation of Guangxi Province (No.
tion. The cumulative error distribution function of this
2018GXNSFAA281164), and the special research project of
deviation is shown in Figure 13. As can be seen from the fig-
Hechi University (No. 2021GCC028, No. 2021XJZD003).
ure, the performance of the cumulative error distribution for
This research was financially supported by the project of
navigation is a little worse than that of the improved WiFi
outstanding thousand young teachers’ training in higher
location fingerprint positioning algorithm. However, the
education institutions of Guangxi, Guangxi Colleges and
probability that the error of navigation is within 0.8 m can
Universities Key Laboratory Breeding Base of System
reach 62%. Therefore, the improved WiFi location finger-
Control and Information Processing.
print localization algorithm is feasible to be applied to the
navigation of the robot.
References
6. Conclusion [1] X. Cai, H. Ning, S. Dhelim et al., “Robot and its living space: a
roadmap for robot development based on the view of living
In this study, our main work is to make corresponding space,” Digital Communications and Networks, vol. 7, no. 4,
improvements to the traditional WiFi location fingerprint pp. 505–517, 2021.
 6302, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2022/8274455 by Readcube (Labtiva Inc.), Wiley Online Library on [08/10/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wireless Communications and Mobile Computing 13

[2] S. Hata, T. Hiroyasu, J. Hayashi, H. Hojoh, and T. Hamada, [19] J. Gao, C. Li, B. Li et al., “Single-event induced failure mode of
“Robot system for cloth handling,” in 2008 34th Annual Con- PWM in DC/DC converter,” Microelectronics Reliability,
ference of IEEE Industrial Electronics, pp. 3449–3454, Orlando, vol. 114, article 113820, 2020.
FL, USA, 2008. [20] M. Oussama, C. Abdelghani, and C. Lakhdar, “Efficiency and
[3] R. Bloss, “Mobile hospital robots cure numerous logistic robustness of type-2 fractional fuzzy PID design using salps
needs,” Industrial Robot: An International Journal, vol. 38, swarm algorithm for a wind turbine control under uncer-
no. 6, pp. 567–571, 2011. tainty,” ISA Transactions, 2021.
[4] R. Pinillos, S. Marcos, R. Feliz, E. Zalama, and J. Gómez-Gar- [21] D. B. Ninh, J. He, V. T. Trung, and D. P. Huy, “An effective
cía-Bermejo, “Long-term assessment of a service robot in a random statistical method for indoor positioning system using
hotel environment,” Robotics and Autonomous Systems, WiFi fingerprinting,” Future Generation Computer Systems,
vol. 79, pp. 40–57, 2016. vol. 109, pp. 238–248, 2020.
[5] Y. Sun, L. Guan, Z. Chang, C. Li, and Y. Gao, “Design of a low- [22] S. Daneshgadeh Çakmakçı, T. Kemmerich, T. Ahmed, and
cost indoor navigation system for food delivery robot based on N. Baykal, “Online DDoS attack detection using Mahalanobis
multi-sensor information fusion,” Sensors, vol. 19, no. 22, distance and kernel-based learning algorithm,” Journal of Net-
p. 4980, 2019. work and Computer Applications, vol. 168, article 102756,
[6] A. Saracoglu and D. U. Sanli, “Accuracy of GPS positioning 2020.
concerning Köppen-Geiger climate classification,” Measure- [23] J. Zhang and H. Mao, “WKNN indoor positioning method
ment, vol. 181, article 109629, 2021. based on spatial feature partition and basketball motion
[7] L. Qin, B. Niu, B. S. Li, X. L. Hu, and Y. X. du, “High precision capture,” Alexandria Engineering Journal, vol. 61, no. 1,
indoor positioning algorithm of single LED lamp based on A- pp. 125–134, 2022.
Bayes,” Optik, vol. 241, article 167190, 2021.
[8] H. Seki, Y. Tanaka, M. Takano, and K. Sasaki, “Positioning sys-
tem for indoor mobile robot using active ultrasonic beacons,”
IFAC Proceedings Volumes, vol. 31, no. 3, pp. 195–200, 1998.
[9] A. Wasik, R. Ventura, J. N. Pereira, P. U. Lima, and
A. Martinoli, “Lidar-based relative position estimation and
tracking for multi-robot systems,” in Robot 2015: Second Ibe-
rian Robotics Conference, Advances in Intelligent Systems
and Computing, Springer, Cham, 2016.
[10] J. Li, X. Zhang, J. Li, Y. Liu, and J. Wang, “Building and opti-
mization of 3D semantic map based on Lidar and camera
fusion,” Neurocomputing, vol. 409, pp. 394–407, 2020.
[11] P. Jiang, L. Chen, H. Guo, M. Yu, and J. Xiong, “Novel indoor
positioning algorithm based on Lidar/inertial measurement
unit integrated system,” International Journal of Advanced
Robotic Systems, vol. 18, no. 2, article 172988142199992, 2021.
[12] Q. Li, J. P. Queralta, T. N. Gia, Z. Zou, and T. Westerlund,
“Multi-sensor fusion for navigation and mapping in autono-
mous vehicles: accurate localization in urban environments,”
Unmanned Systems, vol. 8, no. 3, pp. 229–237, 2020.
[13] G. Li, L. Yu, and S. Fei, “A deep-learning real-time visual
SLAM system based on multi-task feature extraction network
and self-supervised feature points,” Measurement, vol. 168,
article 108403, 2021.
[14] B. Fang, G. Mei, X. Yuan, L. Wang, Z. Wang, and J. Wang,
“Visual SLAM for robot navigation in healthcare facility,” Pat-
tern Recognition, vol. 113, article 107822, 2021.
[15] V. S. Kalogeiton, K. Ioannidis, G. C. Sirakoulis, and E. B. Kos-
matopoulos, “Real-time active SLAM and obstacle avoidance
for an autonomous robot based on stereo vision,” Cybernetics
and Systems, vol. 50, no. 3, pp. 239–260, 2019.
[16] H. Zou, Y. Zhou, J. Yang, and C. J. Spanos, “Towards occupant
activity driven smart buildings via WiFi-enabled IoT devices
and deep learning,” Energy and Buildings, vol. 177, pp. 12–
22, 2018.
[17] Y. Peng, “Research on teaching based on tennis-assisted robot
image recognition,” Microprocessors and Microsystems, vol. 82,
article 103896, 2021.
[18] L. Zhao, S. Qu, and W. Zhang, “Design of multi-channel data
collector for highway tunnel lighting based on STM32 and
Modbus protocol,” Optik, vol. 213, article 164388, 2020.