You are on page 1of 8

2020 IEEE 7th International Conference on Industrial Engineering and Applications

Route Planning for Automatic Indoor Driving of Smart Cars

Sandip Roy Zhiyang Zhang


School of Software Engineering School of Software Engineering
Tongji University Tongji university
Shanghai, China Shanghai, China
e-mail: sandip@tongji.edu.cn e-mail: zhiyang@tongji.edu.cn

Abstract—While various types of traffic navigation maps play scenarios of the automated smart car navigation can be
an important role in our life, the complex layout of large summarized as follows:
number of buildings in the city makes it difficult for people to  Indoor positioning and navigation inside large
find destinations and walking routes indoors. Since GPS does buildings: When entering a large complex building,
not work very accurately inside the buildings, for indoor it is quite a difficult task for people to quickly and
navigation problem we still have to depend on the traditional accurately find the target location. Smart cars can
front desk human resource consultation. In recent years, the help people to plan the route, find the best route, and
rapid development of artificial intelligence technology has direct to the designated location.
made it possible for smart car navigation instead of the manual
 Indoor express sorting and delivery: With the
consultation. Not only that, smart cars can even handle short-
distance goods transportation. Combining indoor positioning
increase of courier temporary storage services these
technology with different video signals or any other non-GPS days it greatly facilitated the time coordination
satellite positioning technology for directing the smart car to between courier delivery and customer pickup, but
the designated position is a hot research direction these days. customers still sometimes feel that this pickup could
In this paper, we mainly study the realization and application have been more convenient. The smart car can
of a smart car in the indoor automated driving system. Our improve this efficiency by sorting those parcels and
proposed smart car, built using the Raspberry Pi 3, uses a then deliver it to the intended person at the entrance
series of famous path planning algorithms such as Dijkstra’s of the customer’s building.
algorithm, Best priority search and A*(A star) algorithm to  Finding lost people: In large public places or
plan the driving path of the car's automatic driving. Finally, buildings, such as airports, train station, shopping
the real-time obstacle avoidance function is also realized by malls etc. when the companion or child is lost, the
using ultrasonic ranging, infrared ranging and pan-tilt camera. smart car can monitor the images in real time by
Several experiments were conducted to compare the efficiency downloading it from the server through the network,
of using different path finding algorithms and the results of identify the furnishings and modeling and thus
using infrared ranging and ultrasonic ranging during obstacle quickly locate the lost children. It can also guide
avoidance. people to the location of their lost baby/companion.
 Public safety: If there is a serious accident inside
Keywords-raspberry Pi; smart car; ultrasonic ranging; polar
the room, the smart car can help people quickly
map; A* algorithm; Dijkstra algorithm; best priority search;
indoor route planning locate themselves and plan their escape route. In
extremely serious incidents, smart cars can help
rescuers locating the victims and thus assist in the
I. INTRODUCTION rescue process.
With the development of human lifestyles and living In this paper, we mainly study the realization and
standards, these days human beings tend to spend more than application of a smart car in the indoor automated driving
70% of their time inside of the large and complex indoor system [4]. After using the horizontal steering angle and
locations. Thus, efficient indoor navigation [1, 2] or route ultrasonic distance measuring device to determine its
planning [14, 18, 20] has become a very important research position in the scene map model, our proposed smart car has
direction these days. While the GPS technology doesn’t implemented few famous path planning algorithms such as
work very accurate for the indoor locations, previous Dijkstra’s algorithm [10, 11, 19], Best priority search [8] and
research works have tackled this problem of indoor A* algorithm [3, 23] to plan the driving path of the car's
navigation using Wi-Fi [9] and sometimes using both Wi-Fi automatic driving. Here the real-time obstacle avoidance [6]
and Bluetooth together (hybrid) [17]. In [21] W. Yuan and function is realized by using ultrasonic ranging [5], infrared
ranging and pan-tilt camera [12]. Our overall system
M. Schneider have introduced 3D route planning for indoor
combines path planning algorithm, embedded system design,
space. When developing our research, we have studied a lot
infrared ranging, ultrasonic ranging, pan-tilt camera, image
of works [16, 22, 24] that have provided a nice overview of
retrieval technology etc. to realize the automatic driving of
different path planning algorithms for robots. The application
smart cars. The camera that is put on the front of the smart

978-1-7281-6785-5/20/$31.00 ©2020 IEEE 743

Authorized licensed use limited to: Universidad Industrial de Santander. Downloaded on May 31,2020 at 22:29:56 UTC from IEEE Xplore. Restrictions apply.
car can take pictures of the target and then the target is intervals into the medium. When that transmitted sound
recognized by the camera. The ultrasonic detecting device wave encounters any object during its propagation then that
measures the distance from the target to the trolley and the sound wave is reflected back to the sensor as an echo. Finally,
angle of the head of the gimbal is used to establish a polar the sensor measures the distance of that target object based
coordinate with the car itself as the origin. Specifically, the on the time spent between transmitting the signal and
main work is as follows: receiving it back. For measuring the distance of any object, it
 Collection of the data, the data source is the indoor does not have to depend on the intensity of the reflected
simulation experiment environment. sound wave and thus it works very good in avoiding any
 Setting up a smart car experimental environment background interferences like light, dust, smoke, mist, vapor,
based on Raspberry Pi 3. lint etc. Pretty much all the materials that are able to reflect
 Calling the map model from the server through sound can be detected using the ultrasonic sensor. Having
WebSocket and returning the target recognition type this great ability that it can also work in dusty air condition
from the server and position coordinates. gives us even more reliability. Only one disadvantage that is
 Determine the position coordinates of the car itself found with using ultrasonic sensors is that they aren’t very
by using the infrared detecting device, the ultrasonic good when defining the edges of an object.
distance measuring device and the pan-tilt camera.
 Implementing Dijkstra’s algorithm, best priority B. Infrared ranging principle
search and A* algorithm for path planning. Infrared sensor is an instrument that uses infrared light to
 Using the infrared detection device, the ultrasonic detect the presence of an object. The infrared ranging sensor
distance measuring device and the pan/tilt camera to uses the principle [13] that depending on the distance of the
realize the automatic obstacle avoidance function. obstacle, the intensity of different reflections of the infrared
This paper is divided into five chapters. The first chapter signal is also different. When the distance is near, the
is the introduction. The development status and research intensity of the reflected light is strong, on the other hand,
background of smart cars are introduced, and the main work when the distance is far, the intensity of the reflected light
of this paper is illustrated. The second chapter is the signal is weak. And that’s how the detection of the distance
theoretical basis for the subsequent implementation, the of the obstacle is performed. The infrared ranging sensor has
principle of ultrasonic and infrared ranging, Dijkstra’s a pair of diodes, infrared signal transmitting diode and the
algorithm, best priority search and the A* algorithm is receiving diodes. The transmitting tube emits an infrared
introduced here. The third chapter is the realization part of signal of a specific frequency and the receiving tube receives
the smart car’s path planning and automatic obstacle the infrared signal of that frequency. When the transmitted
avoidance, showing how to use the theory described in the infrared ray encounters an obstacle, the infrared signal is
second chapter. Here we have introduced the details of each reflected back and received. After receiving and processing,
implementation separately. The fourth chapter is dedicated to it is returned to the smart car host through the digital sensor
the experimentation part. Using the scene map as a model, interface. In our system, the smart car can then use this
according to the realization of the third chapter, the whole infrared return signal to identify changes in the surrounding
system is realized and finally different path planning environment.
algorithms were compared to present the results. The fifth
C. Principle of Web socket
chapter contains the summary and the outlook section. The
research results and significance of this topic are summarized, HTML5 defines the Web socket protocol [15], which
and suggestions for the future development are put forward. saves server resources and bandwidth and enables
communication with the server in real time. Web socket is a
II. REVIEW OF THE RELEVANT THEORETICAL protocol for full-duplex communication over a single TCP
FOUNDATIONS connection. WebSocket makes it easier to exchange data
In the smart car model proposed in this paper, after using between the client and the server, allowing the server to
the horizontal steering angle and the ultrasonic distance actively push data to the client. In the Web socket API, the
measuring device to determine its position in the scene map client/browser can directly create a persistent connection and
model, the smart car uses a series of famous path planning transfer data in both directions. Using HTTP protocol, the
algorithms such as Dijkstra’s algorithm, best-first search and client needs to constantly ask the server if there are any new
A* algorithm to plan the automatic driving of the car. Finally, messages to be received. Web Sockets on the other hand
the real-time obstacle avoidance is done by means of doesn’t require any request from the client side in order to
ultrasonic ranging, infrared ranging and pan/tilt camera. respond. The client can just listen to the server for any new
Overall, the main technical theories used in this paper messages whenever its available. Web sockets work
include infrared ranging, ultrasonic ranging, path finding perfectly for any real-time applications, chat apps, IOT,
algorithms, and WebSocket technology. online multiplayer games etc.

A. Ultrasonic ranging principle D. Path finding Algorithms


The principle of ultrasonic ranging [5, 7] is rather very The path finding algorithm lets us to plan ahead, rather
simple to understand. At first, the ultrasonic sensor emits than waiting until the last minute to find the problem.
short and high frequency sound pulses at regular time Although motion without path searching can work in many

744

Authorized licensed use limited to: Universidad Industrial de Santander. Downloaded on May 31,2020 at 22:29:56 UTC from IEEE Xplore. Restrictions apply.
situations and can be extended to many more situations, but initial point to the target point can be found, as long as all
actually path searching is rather a common approach to solve edges have a non-negative value. In Figure. 3, the pink node
more complex problems. Although Moving a simple object is the initial node, the violet node is the target point, and the
seems to be a very easy job to do but Path finding is actually diamond shaped blue colored area is the area scanned by the
quite complicated. For understanding the reason behind why Dijkstra’s algorithm. The areas with the lightest color are
it is troublesome when it comes to path finding we have to those that are furthest from the initial point, thus forming the
consider the following: frontier of the exploration. One good thing about Dijkstra’s
algorithm is that although it runs slower, but it always does
guarantee that a shortest path will be found.

Figure 1. Concave obstacle path planning.

Figure 3. Dijkstra’s path finding algorithm.

F. Best First Search (BFS) algorithm


The Best First Search (BFS) algorithm operates in a
similar process, except that it can evaluate the cost of any
node to the target point by a heuristic process. Unlike
selecting the nodes closest to the initial node, it selects the
node closest to the target. But BFS cannot guarantee that a
shortest path will be found. However, it is much faster than
the Dijkstra’s algorithm because it uses a heuristic function
Figure 2. Path planning by avoiding obstacle. to quickly direct to the target node. For example, if the target
is located in the south side of the starting point, BFS will
We can consider a simple example (Figure. 1), where the tend to lead the path to the south. However, these two
unit is initially at the bottom of the map and attempts to examples are only the simplest case, where there are no
move to the top. There is nothing in the area scanned by the obstacles in the map, and the shortest path is quite straight.
object (pink part) to act as an obstacle, so it continues to
G. A* algorithm
move up. As it approaches the top, it detects an obstacle and
then changes the direction of its movement. Eventually, it Like other common graph searching algorithms, A*
finds the red path along the U-shaped obstacle. On the other potentially searches for a large area of the graph. Just like
hand, a pathfinder algorithm will scan a larger area (light Dijkstra’s algorithm, A* can also be used to search for the
gray) in (Figure. 2) and thus, it does not allow the unit to go shortest path. Similar to BFS, A* can guide itself through a
to the concave obstacle and finds a shorter path (blue colored heuristic function (the original is heuristic). In the simple
path). However, one can extend a motion algorithm to deal case, it is as fast as BFS. The secret behind its success is that
with the obstacles or to avoid creating concave obstacles it combines the information blocks of the Dijkstra’s
marking the concave exit as dangerous. algorithm (the nodes near the initial point) and the BFS
algorithm (the nodes near the target point). In the standard
E. Dijkstra’s Algorithm language discussing a*, we can consider the equation (1):
The Dijkstra’s algorithm accesses the nodes in the graph
starting from the initial point where the object is located. It F ( n)  g ( n)  h( n) (1)
iteratively checks the unvisited nodes in the node set and
adds the unchecked node closest to the original node to the Here, g(n) represents the cost from the initial node to any
set of nodes that need to be checked. The set of nodes node n, and h(n) represents the heuristic estimation cost from
expands outward from the initial node until it reaches the the node n to the target node.
target node. The Dijkstra’s algorithm uses positive integers
or real numbers as labels which are totally ordered. It can be H. Heuristic Algorithm
generalized to use any labels that are partially ordered. The The heuristic function h(n) tells us the minimum cost
Dijkstra’s algorithm guarantees that a shortest path from the estimation from any node n to the target node when using the

745

Authorized licensed use limited to: Universidad Industrial de Santander. Downloaded on May 31,2020 at 22:29:56 UTC from IEEE Xplore. Restrictions apply.
A* algorithm. Therefore, it is very important to choose a III. IMPLEMENTATION OF AUTOMATIC DRIVING FOR
good heuristic function. Heuristics can control the behavior SMART CARS
of A*. If we consider the general formula in equation (1), in
No doubt, the implementation of different smart cars can
an extreme case, if h(n) is 0, then only g(n) works, and A*
bring great convenience to our life. Here in this work we
evolves into Dijkstra’s algorithm, which guarantees the car
mainly focus on implementation of two aspects of automatic
can find the shortest path. If h(n) is often smaller (or equal)
driving of the smart car, these are:
than the actual cost of moving from n to the target, A* is
 Implementation of route planning
guaranteed to find the shortest path. The smaller the value of
 Real-time obstacle avoidance function
h(n) is, the more nodes in A* needs to be visited which
The route planning is implemented based on the path
makes the operation very slow. Here we can consider few
finding algorithms and the car's own positioning system, and
scenarios:
the real-time obstacle avoidance function is realized with the
 If h(n) is exactly equal to the cost of moving from n
help of the infrared sensor, the ultrasonic sensor and the
to the target, then A* will only find the best path and
pan/tilt camera.
not extend the other. Although this can't happen in
all situations, one can still be in some special cases. A. System Architecture Overview
Let them be exactly equal (h(n) is exactly equal to
the actual value). As long as the perfect information
is provided, A* will run perfectly.
 If h(n) is sometimes higher than the actual cost of
moving from n to the target, A* cannot guarantee
that a shortest path will be found, but it will run
faster.
 In another extreme case, if h(n) is much larger than
g(n), then only h(n) works and A* evolves into the
BFS algorithm.
So here we have found a very interesting situation, we
can actually decide what do we want from A*. In the ideal
case we want to find the shortest path as quickly as possible.
If our goal is too low, we will still get the shortest path, but
Figure 4. Smart car.
the speed will be much slower. On the other hand, if our goal
is too high then might have to give up the idea of finding the Due to the rapid development of the Raspberry Pi
shortest path. Academically, if the heuristic function value is microcomputer in recent years, considering its reach
an underestimation of the actual cost, the A* algorithm is functionality and better scalability, we have also chosen
called the simple A algorithm (Simply A). However, here we Raspberry Pi 3 to build our smart car (Figure. 4). While
continue to call it A* because it is the same when it comes to another open source electronic platform, Arduino is also very
the implementation. easy to develop, but the open source community is not as
We have seen that it is a crucial decision to make during active as the Raspberry Pi's open source community and does
choosing between speed and accuracy. A*'s ability to change not have higher scalability. Therefore, after overall
its own behavior is based on heuristic cost functions, which comparison, we finally chose to use Raspberry Pi 3 [9].
are very useful in games. A compromise between speed and During implementation of our overall design (Figure 5) we
precision will make the game run faster. In many games, we focus on the following four points:
don't really need to get the best path, just the approximation  Positioning of the trolley in the experimental scene
is enough. What we need depends on what happens in the map
game or how fast the machine is running the game. The  Path planning of the car to the target
choice between speed and accuracy is not global. In certain  Automatically generate the car driving code
areas of the map, accuracy is important, we can use dynamic according to the planning path
selection based on this. For example, suppose we might stop  Real-time obstacle avoidance function
recalculating a path or changing direction at some point, then The car is built using the Raspberry Pi 3 and it uses
it is more important to choose a good path near the current Python scripts to control motor drives, pan-tilt camera and
position. So why do we need to bother with the accuracy of infrared sensors. The experimental scene map model is
subsequent paths? For a safe area on the map, the shortest provided from the already conducted experimental results
path may not be important, but when escaping from an stored in a server. The trolley determines the position of
enemy village, safety and speed are the most important. Here, itself by detecting the distance of the surrounding objects and
in the safe area, we can consider not finding the exact the angle of rotation of the steering gear by using infrared
shortest path and taking the approximate path making the rays. The path planning of the car to the target uses three
pathfinding process very fast, but in the dangerous area, both algorithms to compare the results, namely, Dijkstra’s
the safety and the speed of escape are equally important algorithm, Best priority search and A* algorithm.
factors to consider.

746

Authorized licensed use limited to: Universidad Industrial de Santander. Downloaded on May 31,2020 at 22:29:56 UTC from IEEE Xplore. Restrictions apply.
Figure 7. Infrared ranging principle.

However, both of these ranging methods have congenital


defects.
Figure 5. Overall system architecture.

t1  t1  t0 , L   c  t   / 2 (3)


B. Realization of ranging of ultrasound detection device
It is only necessary to place the ultrasonic ranging Here, c represents the speed of light, the time of
module at the front of the trolley and access the Raspberry transmitting the infrared ray is represented by t0 , the time
Pi’s input and output port. Each time the ultrasonic wave is
received after getting reflected from the obstacle, the of receiving back that reflected infrared ray is t1 and t1
distance of the obstacle from the car can be calculated (figure
6) according to the following formula (2). represents the time interval between transmitting the sound
wave and receiving back the echo after its reflection. We can
calculate the distance from the obstacle in front of the car to
(2)
itself, L according to the formula (3)
D. Implementation of the Rotating Device Based on
Double-Steering Machine
As shown in figure. 8, each time the pan-tilt horizontal
servo scans from left to right, the orientation of the camera,
infrared detector and ultrasonic device is consistent with the
front of the servo. In the direction of the rotation ∆d, if the
camera is shooting when the object is matched with the
Figure 6. Ultrasound ranging principle. object in the map model, it is determined that the closest
object in the direction is the matched object
Here, the time of transmitting the ultrasonic wave is
represented by t0 , the time of receiving back the reflected
ultrasonic wave is t1 and t represents the time interval
between transmitting the sound wave and receiving back the
echo after its reflection. The measured distance is L and the
speed of sound is expressed by  . During conducting the
experiment, under standard atmospheric pressure and 15 ° C
temperature condition the speed of sound is taken as
  340 m/s. After each time the ultrasonic wave is
received, we calculate the distance from the obstacle in front
of the car to itself according to formula (2).
C. Realization of Ranging of Infrared Detection Device Figure 8. Implementation of the rotating device.
Just like the ultrasonic sensor it is only necessary to place
the infrared ranging module at the front of the car and access
E. Conversion of Coordinate System and Determination of
the distance from the raspberry pie to the obstacle in front of
the car. The output port can calculate the small time Coordinates of smart car
according to the formula (equation. 3) after acquiring the According to L and d (Figure. 9), taking obstacle
time data of transmitting infrared rays and receiving back B as the poles and the counter clock wise direction as the
that reflected infrared ray, shown in figure 7 positive direction, it is possible to establish the polar
coordinate system with the horizontal steering gear. The

747

Authorized licensed use limited to: Universidad Industrial de Santander. Downloaded on May 31,2020 at 22:29:56 UTC from IEEE Xplore. Restrictions apply.
coordinates of the obstacle B are    d , L  with respect B. Hardware Environment
to the coordinate system and the horizontal steering gear as TABLE I. HARDWARE ENVIRONMENT
the pole (figure 9). Then, according to the orthogonal angle
coordinates of the obstacle B in the map model the cartesian Raspberry Pi 3 model B+ Quad core 1.2 GHz 64-bit CPU
coordinates of the trolley are inferred. According to the
transformation and correspondence of the coordinates, the CPU 8 GB 1600 MHz DDR3

coordinates  xcar , ycar  of the cart in the map model are RAM 16GB
determined, where
WLAN BCM43438 wireless LAN

xcar  xb  L  sin   d  , Bluetooth Low energy (BLE) on board

ycar  yb  L  cos(  d )
Ethernet 100 Base Ethernet

GPIO 40 pins extended GPIO

USB 4 USB 2 ports

Video and Sounds 4 pole stereo output

Screen output Full size HDMI

CSI camera port For connecting Raspberry pi


camera

C. Software Environment

Figure 9. Conversion of coordinate system and determination of TABLE II. SOFTWARE ENVIRONMENT
coordinates.

IV. EXPERIMENT Operating system Linux Debian

Python environment Python 2.7.5


A. Experimental Process and Comparison of the Results
The experiment process is to first take a picture of the OpenCV environment OpenCV 3.4
target by using the camera of the car and then transmitting
the target photo to the server. The server program then Development platform PyCharm
recognizes the target and returns the target position and the
map model. After the car locates its position automatically it
then starts to plan the path towards the target. Furthermore,
during the whole process it also avoids any obstacles on the
way of its journey. During experimentation a large number
of different scene maps were tested. The experiments
compared the performances of the three path finding
algorithms (figure 11), and finally chose the A* algorithm as
the best path finding algorithm considering the path length
and calculation speed. When Comparing between two
distance measuring devices, ultrasonic and infrared sensors
(figure 10), the ultrasonic sensor is found to be working
better for the smaller distance (less than 6 meters), but
infrared light sensor worked better for the larger distance
(more than 6 meters). On the other hand, the infrared anti-
interference is poor and the result is very easy to vary with Figure 10. Obstacle avoidance performance.
light and high temperature interference.

748

Authorized licensed use limited to: Universidad Industrial de Santander. Downloaded on May 31,2020 at 22:29:56 UTC from IEEE Xplore. Restrictions apply.
REFERENCES
[1] J. Dong, Y. Xiao, Z. Ou, Y. Cui, and A. Ylä-Jääski, "Indoor tracking
using crowdsourced maps," in Proceedings of the 15th International
Conference on Information Processing in Sensor Networks, 2016, p. 5.
[2] J. Dong, Y. Xiao, M. Noreikis, Z. Ou, and A. Ylä-Jääski, "imoon:
Using smartphones for image-based indoor navigation," in
Proceedings of the 13th ACM Conference on Embedded Networked
Sensor Systems, 2015, pp. 85-97.
[3] F. Duchoň, A. Babinec, M. Kajan, P. Beňo, M. Florek, T. Fico, et al.,
"Path planning with modified a star algorithm for a mobile robot,"
Procedia Engineering, vol. 96, pp. 59-69, 2014.
Figure 11. path finding algorithms performance. [4] T. Tao, "Fuzzy automatic driving system for a robot car," in 2014
International Conference on Fuzzy Theory and Its Applications
(iFUZZY2014), 2014, pp. 132-137.
V. CONCLUSION AND FUTURE WORK [5] J. Borenstein and Y. Koren, "Obstacle avoidance with ultrasonic
Here in this work we have successfully completed the sensors," IEEE Journal on Robotics and Automation, vol. 4, pp. 213-
218, 1988.
positioning of the smart car, route planning and automatic
[6] J. Borenstein and Y. Koren, "Real-time obstacle avoidance for fast
obstacle avoidance. The experiment is based on the raspberry mobile robots," IEEE Transactions on systems, Man, and Cybernetics,
pie-based smart car plus the ultrasonic and infrared detecting vol. 19, pp. 1179-1187, 1989.
device and the cloud. Few famous path finding algorithms [7] J. L. Crowley, "World modeling and position estimation for a mobile
are used here in the previously trained scene map model to robot using ultrasonic ranging," in Proceedings, 1989 International
realize the route planning of the automatic car driving. Conference on Robotics and Automation, 1989, pp. 674-680.
Finally, the real-time obstacle avoidance function is applied [8] R. Dechter and J. Pearl, "Generalized best-first search strategies and
by using the pan-tilt camera and the ultrasonic detecting the optimality of A," Journal of the ACM (JACM), vol. 32, pp. 505-
device to avoid pedestrians or any other unknown mobile 536, 1985.
targets walking between the targets. In addition, here we [9] D. Han, S. Jung, M. Lee, and G. Yoon, "Building a practical Wi-Fi-
based indoor navigation system," IEEE Pervasive Computing, vol. 13,
have used Python and WebSocket technology to achieve car- pp. 72-79, 2014.
to-server communication, showing high efficiency in writing [10] Y. Y. G. Jianya, "An Efficient Implementation of Shortest Path
code and in actual use. The user only needs to take the photo Algorithm Based on Dijkstra Algorithm [J]," Journal of Wuhan
of the target using the camera of the car and no other further Technical University of Surveying and Mapping (Wtusm), vol. 3,
operation is required. Everything else is independently 1999.
completed by the smart car, which greatly simplifies the [11] H. I. Kang, B. Lee, and K. Kim, "Path planning algorithm using the
process and service of many manual consultations for the particle swarm optimization and the improved Dijkstra algorithm," in
2008 IEEE Pacific-Asia Workshop on Computational Intelligence
production and life of the society. and Industrial Application, 2008, pp. 1002-1004.
A. Future work Recommendation [12] K. Kurihara, S. i. Hoshino, K. Yamane, and Y. Nakamura, "Optical
motion capture system with pan-t mera tracking and real time data
Now-a-days, although the research on smart cars are processing," in Proceedings 2002 IEEE International Conference on
endless, there is no mature car that is popular among the Robotics and Automation (Cat. No. 02CH37292), 2002, pp. 1241-
public, the available ones are really expensive too. We feel 1248.
that with the continuous development of artificial [13] E. Lavarec and L. Tremel, "Method and device for obstacle detection
intelligence, the innovation of embedded robots, embedded and distance measurement by infrared radiation," ed: Google Patents,
2004.
car chips and hardware technology can definitely bring
[14] D. Mandloi and J.-C. Thill, "Object-oriented data modeling of an
breakthroughs in quality and efficiency for the realization of indoor/outdoor urban transportation network and route planning
smart cars. In the present era of big data, positioning services analysis," in Geospatial Analysis and Modelling of Urban Structure
like Google or Baidu maps have a huge amount of data for and Dynamics, ed: Springer, 2010, pp. 197-220.
deep learning, which can support more accuracy. At that [15] V. Pimentel and B. G. Nickerson, "Communicating and displaying
time one can choose more efficient deep learning method real-time data with websocket," IEEE Internet Computing, vol. 16, pp.
than the classic algorithm to complete the path planning. In 45-53, 2012.
addition, obstacle avoidance technology can achieve more [16] N. Sariff and N. Buniyamin, "An overview of autonomous mobile
precise positioning by combining multiple position signals, robot path planning algorithms," in 2006 4th Student Conference on
Research and Development, 2006, pp. 183-188.
such as Wi-Fi signals or communication signals. The
[17] H.-K. Su, Z.-X. Lıao, C.-H. Lin, and T.-M. Lin, "A hybrid indoor-
combination and correction of multiple signals can greatly position mechanism based on bluetooth and WiFi communications for
improve the accuracy in practical use. It is also expected that smart mobile devices," in 2015 International Symposium on
more efficient computer vision algorithms will allow the Bioelectronics and Bioinformatics (ISBB), 2015, pp. 188-191.
camera to do better obstacle avoidance. [18] T.-A. Teo and K.-H. Cho, "BIM-oriented indoor network model for
indoor and outdoor combined route planning," Advanced Engineering
ACKNOWLEDGMENT Informatics, vol. 30, pp. 268-282, 2016.
[19] H. Wang, Y. Yu, and Q. Yuan, "Application of Dijkstra algorithm in
We want to thank Professor Weixiong Rao of Tongji robot path-planning," in 2011 second international conference on
university for his overall supervision and guidance. mechanic automation and control engineering, 2011, pp. 1067-1069.

749

Authorized licensed use limited to: Universidad Industrial de Santander. Downloaded on May 31,2020 at 22:29:56 UTC from IEEE Xplore. Restrictions apply.
[20] J. Yu and S. M. LaValle, "Planning optimal paths for multiple robots [22] G. Zhang, X. Hu, J. Chai, L. Zhao, and T. Yu, "Summary of path
on graphs," in 2013 IEEE International Conference on Robotics and planning algorithm and its application," Mod. Mach, vol. 5, pp. 85-90,
Automation, 2013, pp. 3612-3617. 2011.
[21] W. Yuan and M. Schneider, "Supporting 3D route planning in indoor [23] S.-p. ZHANG, X.-k. WANG, and J. DUAN, "Application of A-star
space based on the LEGO representation," in Proceedings of the 2nd Algorithm in Mobile Robot Path Planning," Mechanical Engineering
ACM SIGSPATIAL international workshop on indoor spatial & Automation, p. 59, 2012.
awareness, 2010, pp. 16-23. [24] D.-q. Zhu and M.-z. Yan, "Survey on technology of mobile robot path
planning," Control and Decision, vol. 25, pp. 961-967, 2010.

750

Authorized licensed use limited to: Universidad Industrial de Santander. Downloaded on May 31,2020 at 22:29:56 UTC from IEEE Xplore. Restrictions apply.

You might also like