2015 3rd International Conference on Signal Processing, Communication and Networking (ICSCN)

Path Planning Algorithm for Autonomous Mobile

Robot in Dynamic Environment
M.S.Ganeshmurthy I
Dr.G.R.Suresh 2
PG scholar, Professor,
M.E Embedded System Technologies, Department of Electronics and Communication,
Easwari Engineering College. Easwari Engineering College,
Chennai, Tamilnadu, India. Chennai, Tamilnadu, India.
Email:murthymsganesh@gmail.com Email:sureshgr@rediffinail.com

Therefore, research on path planning algorithms to make

Abstract- Mobile robots are mostly used in many industrial the robot move from a specified start point to the termination
and military applications. Researches in the path planning is one
point without colliding with obstacles is a fundamental
of the most important aspects in mobile robot researches. Path
requirement for the mobile robot safety in such environments.
planning for a mobile robot need to find a collision free path
through the robot's environment with obstacles from the
Hence, to reduce the energy consumption, processing time and
specified start location to a desired(target) goal location while communication delay the planned path is naturally required to
satisfying certain optimization conditions. Existing path planning be optimal with the shortest length.
methods like graphical methods such as visibility graph, the The complete information of a dynamic environment will
potential field and the cell decomposition are designed for static change along with the movement of obstacles, the complexity
environments, in which there are stationary obstacles. In and uncertainty of the path planning increased greatly in
practical systems such as Marine Science Research, Robots in dynamic environments.
Industry, and military combat applications, robots usually face
dynamic environments where both moving and stationary
III. EXISTING PATH PLANNING ALGORITHMS AND ApPROACHES
obstacles exist. This project proposes a heuristic based method to
search the feasible initial path efficiently. The heuristic based
method is then combined into the simulated annealing algorithm Traditional static path planning methods, such as
based approach for dynamic robot path planning. Therefore the Visibility Graph [8], Voronoi diagrams [2] and Grids [10], are
quality of the solution is characterised by the length of the not suitable for planning path in dynamic environments
planned path and it is improved with the combined heuristic Recently genetic algorithm based navigation method [6] is
method in the simulated annealing based approach for both introduced to deal with the path planning problem in an
runtime and offline path planning. environment with moving obstacles. However, this method
has drawbacks if it is a local minimal results may occur in
Keywords- path planning, static environment, dynamic
environment, simulated annealing. some cases and the calculating time for finding the first
feasible path increases greatly while the number of the
I. INTRODUCTION
obstacles in the environment increases. The resultant that is
feasible path convergence rate is high using genetic
algorithms, making the method lose the mechanism for
This project develops as well as implements a simulated
annealing algorithm based approach with a heuristic search exploration. If this method has a low convergence rate, it
method for path planning for a mobile robot in dynamic could raise the mutation probability and reduce the crossover
environments. The approach uses the vertices of the static and probability. Therefore, an improved method for path planning
dynamic obstacles as search space. Using the heuristic search in dynamic environment needs to be designed
method initial feasible path for the mobile robot in the
environment is identified. Then, the approach applies the SA The A* algorithm [3] is another path planning method to
algorithm to find the optimal path for the robot. When a help the robot to find the optimal path in grid decomposed
moving obstacle is detected with possible collision with the static grid maps. The environment with free space and
robot, this two-step planning process is activated for dynamic obstacles is presented by a set of uniformed regular grids. The
path planning. A-star uses heuristic based Dijkstra algorithm to obtain the
II. PROBLEM STATEMENT optimal result for the robot. The drawback is that the A* uses
uniformed grids representation which must allocate specific
At present, research on various algorithms for mobile large amounts of memory for regions that may never be
robot path planning is a hot topic. Mobile robots are mostly traversed or may not contain the obstacles. This drawback
used in many hazardous industrial fields where there may be may affect the efficiency of the method. Quad tree method
dangerous for people, such as aerospace research, the nuclear may be used to enhance the efficiency of the method, but it
industry, and mining industry. To find a safe path in the sacrificed the optimality for the efficiency. The above
dangerous environment for the mobile robot is a basic and drawbacks also happen in the D* method [3] which is the
essential requirement for the success of any mobile robotic dynamic version of the A* method. The D* helps robot to find
systems. the optimal path in uncertain environments. Besides, in
dynamic environments with moving objects, the environment

978-1-4673-6823-0/15/$3l.00 ©2015 IEEE

 2015 3rd International Conference on Signal Processing, Communication and Networking (ICSCN)

changes the planning problem from fixed obstacles to moving A. Structure and Assumptions a/the Method
ones; changes the problem from the geometric one to a
dynamic one; equally the environment changes the problem First, as in the development of many existing methods,
from a deterministic problem to a stochastic problem. some assumptions are made before designing the algorithm.
The genetic algorithm based approach [11] has been
developed for robot path planning in dynamic environment 1) There are stationary and moving obstacles in the
and the efficiency will not sufficient for large-scale path dynamic environment.
planning problems. Therefore, more efficient path planning 2) As a traditional assumption, both the stationary and
dynamic obstacles are bounding polygons.
methods in dynamic obstacle environments need to be
3) The search space is the vertices of the static obstacles
developed for adapting the development of robots research plus the vertices of the detected moving obstacles.
To develop and implement a algorithm based simulated 4) The movement trajectory of the dynamic obstacles is
annealing approach [4] for path planning for mobile robots in constituted by a series of line segments, and is constrained by
dynamic environments; the approach is expected to be able to a straight line.
determine or calculate the optimal feasible path for a robot in 5) Any changes in motion parameter of the dynamic
an environment with dynamic obstacles and to give a result obstacle are immediately available to the robot, if the obstacle
comparable or better than those results in the previous is in the range of the sensor. The time of the robot obtaining
approach the moving parameter (such as speed and direction) is
minimum that does not affect the path planning calculation.
6) The robot could change its speed and direction at any
IV. PROPOSED METHODOLOGy-SIMULATED
time.
ANNEALING APPROACH 7) After the sensor of the robot acquires the motion
The simulated annealing algorithm is a generic information of the moving obstacle. It is assumed that a
probabilistic meta-algorithm process for the global separated module is used to monitor the movement of the
optimization problem like locating a good approximation obstacles for the robot. The module will predict the upcoming
result to the global optimum solution of a given function in a movement for the robot based on the previous movement of
large search space. It is applied in many industrial fields. The the obstacle. It will report any change of direction and speed
application of the simulated annealing algorithm is for the of the obstacle. This module is separated from the robot,
Traveling Salesman Problem (TSP). The TSP problem is that which can be simply implemented by using extern sensors in
real robot implementation.
given a number of cities and the cost of traveling form one
The designed structure of the simulated annealing
city to another, to get the least cost round trip to visit each city algorithm in dynamic obstacles environments is divided into
exactly once and then return to the start city two stages: off-line calculation for stationary obstacles, and
online calculation.

Stage 1: Offline calculation for the path on stationary

obstacle based on the vertices of the static obstacles. The
simulated annealing based approach firstly calculates the
optimized path for the mobile robot base on the positions of
the stationary obstacles.
Once the path is ready, the robot starts to travel through
the stationary obstacles with the sensor that could detect the
360 rotate direction of the robot when moving obstacles are
detected.

Figure 1: State Transformation diagram of simulated

annealing approach

The aims of this project is to implement a simulated

annealing based approach into robot path planning in a
dynamic environment. The simulated annealing approach is
expected can quickly determine the optimal feasible path for
robot in the environment with moving obstacles. The research Figure 2: Scenario 1 of moving obstacle
uses the mathematic software MATLAB & VREP to simulate
the static environment, real world robots and sensors. The
research then implements the simulated annealing path
planning method in the environment.
 2015 3rd International Conference on Signal Processing, Communication and Networking (ICSCN)

Stage 2: Online path calculation once the moving obstacles

are detected by the sensor. As moving obstacle comes under
the
Target
Moving
Obstacle /..-

detection range of the robot, the sensor will detects the

dynamic obstacle and the robot acquire the moving
YES
information like speed and moving direction of the obstacle
n =n+l
from the sensor of the robot.
Then the robot uses the motion information to calculate
the possibility of the moving obstacle clashing with the robot.
If it is calculated that the moving obstacle won't collide with
the robot, the robot will use its current path plan to travel
through the map. As Figure 3.1 shows, the sensor of robot
calculates out that the L shaped moving obstacle will not
collide with the robot. Therefore, the robot will not change the
current planned path. If the moving obstacle will clash with
the robot as Figure 1 shows, the robot will activate the
algorithm again to calculate a new path for the robot.
Figure 3: Scenario 2 of moving obstacle

B. Simulated Annealing Algorithm and Expressions

Traditionally, the path length Ef is the evaluation criteria for the
solution derived from the algorithm. The shortest path will be the
better solution. A feasible solution is been expressed by a series of
vertices linking the start point through to the end point. Each vertex
of the obstacle has its specific series number. The path for the robot is
represented by a sequence of vertex numbers. Thus, the feasible
solution Xis given by:
Figure 4:
X= {Vst art, Vstart+l, Vstart+2...Vend-l, Vend}; Procedure of simulated annealing approach for path planning [4].

The pseudo-code this algorithm is as follows: 1. T= Tinit;

2. while (T>Tterminate)
Step 1: Set an initial temperature T; 3. randomly generate one feasible solution Xs;
Step 2: Generate a initial path order randomly from the start to 4. evaluate Xs, Ef= f(Xs);
5. count= 1;
the destination and calculate the path length L(initial);
6. while (count < Threshold)
Step 3: Generate a new path which also from the start to the 7. generate a new feasible solution Xn base on Xs;
end 8. evaluate Xn,
and calculate path length L(new); 9. En= f(Xn);
Step 4: If L(new) < L(initial), accept new path. Else possibly 10. if f(Xn) < f(Xs) Xs= Xn;
accept new order according to some scheduling. 11. else if rand (1) < (exp«f(Xs)- f(Xn))/T) Xs= Xn;
Step 5: Repeat step 3 and step 4 until the temperature gets 12. count=count + 1;
down. 13. end while
14. T= cool rate * T;
Step 6: Down the temperature and return to step 1
15. update Xs at each reduction of temperature T
16. end while
The following are pseudo-codes for the simulated
annealing algorithm. Xs is the approximate optimal solution for the robot.
 2015 3rd International Conference on Signal Processing, Communication and Networking (ICSCN)

C. Initial Feasible Path 5. IP = {IP,Vs}; endif

6. if Vs destination point; break;
=
It is known that after each degree reduction in the initial
7. else
temperature T, a new feasible solution Xn is selected in every
8. continue;
new round. It is most essential for the algorithm to quickly and 9. endif
correctly generate a random feasible path in each round. For 10. endwhile.
this purpose, the edges of the static obstacles should be D. Online Path Planning
specified first. An edge is any straight-line between two points
on the edge of or within the obstacles. Any path section A robot uses the path generated by offline planning to
calculated crosses the edge is defined as an invalid path. In the travel through the stationary obstacles, the online path planner
proposed program there will be a separate array is used to is triggered automatically to calculate a different optimal path
store the edges of the map. when a moving obstacle is detected. The sensor range of the
robot can be simulated or assumed to be fixed value. The
distances between the robot and every vertices of the moving
obstacle real-timely get updated. If the distance between the
robot and any vertex of the moving obstacle is shorter than the
fixed value, it can be defined that the moving obstacle enters
the range of the sensor, and can be detected by the sensor of
the robot. Then, the separate module as described in
NO assumption (7) (Section A) can monitor the trajectory and the
shape of the moving obstacle. The moving information of the
dynamic obstacles collected by the sensor of the robot which
includes speed and moving direction of the mobile robot.
NO
After the robot combines the information of the dynamic
YES
obstacle with the moving parameter of itself, the robot can
calculate the possibility of collision with the moving obstacles.

Target

... ...
Moving

,,,, ,,
Obstacles
" ", ,

YES C'Clj
:\:/t!
/
, \,
..
Static
Obstacles 2

I
Robot
-
./,/---
-
0Static
Obstacles 1

Figure 5: Initial Feasible Path Planning

At the initial stage of the program a vertex is chosen Figure 6: Online Path Planning
randomly on the map except dynamic objects, starting point
and end point, then use a strait-line to connect the start point The simulated annealing optimization algorithm for
and the selected vertex. If the path line intersects with any of finding optimal path is called when it is calculated that the
the edges of obstacle in the map, the path is considered as robot will collide with the moving obstacle if there is no
invalid path. Another vertex is randomly selected again for change of movement is made in the future. The simulated
testing. If the straight line to the selected vertex does not annealing algorithm will be activated with the updated search
intersect with any edges of obstacles in the map, the straight space. The search space for the algorithm will be updated. The
line is recognized as part of a valid path being calculated. current status of the robot, the current information of the
Then add the vertex into the path. After that, start from the dynamic obstacle, the location information of the dynamic
selected vertex to fmd next valid vertex. This procedure is obstacles that may cause the collision and the location
repeated until the end point is selected and also the path line to information of the static obstacles are combined to a new
the end point is a valid path, for example, The path to the end search space for the robot. In a realistic case, the planning time
point does not intersects with any edges. The following should be short enough for the robot to implement motion
pseudo-codes and Figure 4 illustrate the process of the initial changes to avoid collision, With the search space becomes
path selection steps for initial Path selection process [9]. larger, then the time required planning a path becomes longer.
1. initial path IP= {start}; while (I) Therefore, an efficient algorithm is required that its planning
2. randomly select Vs; time is relatively short for the robot to change its directions to
3. if path{IPend, Vs} intersects any obstacles; continue; avoid the collision with the moving obstacles, which is one
4. else main aim of my project.
 2015 3rd International Conference on Signal Processing, Communication and Networking (ICSCN)

---
II
� _ [J
V. SIMULATION RESULTS & DISCUSSIONS
x

File Edit v-_ In!iftt Too" Desktop Window H�p

0 ,31'''''\ � " 'H) ® � '" � 0 [1 Ull

One of the path planning algorithm (e.g.: A-star 11

Algorithm) for static environment is simulated using 10

MATLAB and the dynamic path planning is demonstrated in

virtual 3d robot environment (VREP). A* is a best-first search
and fmds the least-cost path from a given initial node to one
goal node that is out of one or more possible goals.
It uses a distance as well as cost heuristic function
usually denoted by f(x) to determine the order in which
the search visits nodes in the tree. The distance and cost
heuristic is a sum of two functions: Please Sele<:t Ihe VehICle Inilial posrtlon

The path-cost function, which is the cost from the

starting node to the current node usually denoted by g(x) and Figure 8: Robot calculates the path from source to destination
the "heuristic estimate" of the distance to the goal usually
denoted by h(x). B. Simulationfor Dynamic Environment

A. Simulationfor initialfeasible path (Static

Environment)

The Algorithm for initial feasible path like static

environment is simulated as three parameters vehicle, target
and obstacles. The algorithm calculate the shortest path by
avoiding the obstacles to reach the destination the problem is
that the vehicle is need to know the position of the obstacle
initially before calculating the path.

Figure 9: Initial Feasible Path Selection process

In VREP the Red blocks are the moving obstacles in which it
move to new position at a certain time so it is dynamic. The
" ..... ;. .... ; ....., .....; .. .. ; .....;...... ;... ...; ......;..... ;
robot calculates the initial feasible path and when the robot
o 0 0 0
encounter the dynamic obstacles it reroute the path to
o 0 0 r-vet 0
destination.

o 0 0 VI. IMPLEMENTATION OF PATH PLANNING ON MOBILE ROBOT

The implementation is carried out in a mobile robot that is

'1 2 J 4 5 , 7 • 9
o

10
0

I'
driven by an arm controller. The robot can be controlled
manually as well as automatically plan the path in unknown
dynamic environment.
The Hardware model of the Mobile robot vehicle is shown
in Figure 10
Figure 7: 2D Static Environment with obstacles
Figure 10: Hardware model of the Mobile robot
In Figure 7 the Initial Position of the vehicle, Target and
obstacles are defined in 2D static Grid map. The Vehicle must
know the position of the obstacles to avoid during the path to
destination.
In Figure 9. The vehicle finds the shortest path using A­
star algorithm to reach the destination using Djakarta's and
Heuristic approach. A* is one of the best-first search and fmds
the least cost effective path from a given initial node to one
goal node that is out of one or more possible goals.
The key to determining which grid to use when figuring
out the path is the following equation
f = g + h The Mobile robot is controlled by a remote device using a
web page control like the command is passed to the server
g -> The movement cost is to move from the starting point
configured SBC on the Mobile robot using a webpage.
to a given square on the grid
h -> the estimated cost to move from that given square on The control is either automatic or manual the webpage is
the grid to the fmal destination configure like it is shown in Figure 11
 2015 3rd International Conference on Signal Processing, Communication and Networking (ICSCN)

for robots, the approach searches the initial feasible path for
robot in dynamic environment which contains static and
dynamic obstacles, and then uses simulated annealing
algorithm to obtain the optimal path for the robot in the
dynamic map.
The Novelty of the project is that the Mobile robot
can be a controlled manually and also operated autonomously
from anywhere on the network. The static and dynamic path
Figure 11: Webpage control from the Remote Device planning algorithm are studied and one of the path planning
algorithm (e.g.: A-star Algorithm) for static environment is
The webpage is hosted in the Single Board computer and simulated using MATLAB and the dynamic path planning is
can be viewed anywhere on the network so the robot can be demonstrated in virtual 3d robot environment (VREP) as the
controlled anywhere from the network. This project can be same to be implemented in hardware.
used in applications like Spy Robots, getting the remote
hostage locations in unknown terrain References
The Dynamic environment is sensed using sensors and
sophiscated image processing and the path of the robot is [I]. Ayers .1 [July 1994] "Underwater Walking", Arthropod Structure
determined autonomously or manually from the remote & Development, vol. 33, pages 347-360.
device. [2]. Cook .1 [July 1992] "Adding Intelligence to Robot Arm Path
Planning Using a Graph-Match Analogical Reasoning System"
IEEE International Conference on Intelligent Robots and Systems,
vol. I, pages 657-663.
[3]. Dong Jin Seo, Jongwoo Kim [2013] "Development of
Autonomous Navigation System for an Indoor Service Robot
Application" IEEE Transactions.
[4]. Introduction of Simulated Annealing Algorithm, Internet Source,
-
MobIle Robot <http://en.wikipedia.org/wiki/Simulated_annealing >, From
Wikipedia, the free encyclopedia.
[5]. Kalra .N, D. Ferguson and A. Stentz [Feb 2009] "Incremental
Ta'1et
Reconstruction of Generalized Voronoi Diagrams on Grids",
Robotics and Autonomous Systems, vo1.57, pages 123 - 128.
[6]. Lin Lei, Houjun Wang and Qinsong Wussued [2006] "Improved
Figure 12: Real time scenario of the Environment with Genetic Algorithms Based Path planning of Mobile Robot under
Obstacles Dynamic Unknown Environment", IEEE Transactions.
[7]. Liping Cheng and Chuanxi Liu. Bo Yan [2014] "Improved
Hierarchical A-star Algorithm for Optimal Parking Path Planning
The Mobile robot moves through the obstacle filled of the Large Parking Lot" IEEE Transactions. problems in terrain
environment to find the target. The robot is equipped with a navigation", Artificial Intelligence, vol. 37, pages 171-201.
5MP camera to identifY the target and the environment [8]. Mitchell .H [Dec 1988] "An algorithmic approach to some
parameters and also the mobile robot monitors the unknown O'Dunlaing .C, and C. Yap [1982] "A retraction method for
planning the motion of a disc", Journal of Algorithms, vol. 6,
terrain for the anomalies. pages 104-111.
[9]. Park .1, D. Nam and C. Park [Apr 1997] "An Empirical
Comparison of Simulated Annealing and Genetic Algorithms on
NK Fitness Landscapes", Proceeds of IEEE the Sixth International
Conference on Evolutionary Computation, pages 147-151.
[10]. Payton .0, 1. Rosenblatt and D. Keirsey [1993] "Grid-based
mapping for autonomous mobile robot", Robotics and
Autonomous Systems, vol. 11, pages 13-21.
[11]. Peng .Y and W. Wei [Nov 2006] "A New Trajectory Planning

E��"--
Method of Redundant Manipulator Based on Adaptive
•• ,. !:to 10. 11111 1-0. ". _
. .

Simulated Annealing Genetic Algorithm",

JllIII _ l• •• ItlllI U.
[12]. Proceedings of IEEE the Sixth International Conference on
Intelligent System Design and Applications, vol. 1, pages 262-266.
Figure 13: Real-time Temperature and Humidity Analysis [13]. Rathbun .0 and S. Kragelund [Oct 2002] "An Evolution Based
Path Planning Algorithm for Autonomous Motion of a UAV
Through Uncertain Environments", IEEE International Conference
A Graph is plotted between Temperature and Humidity on Digital Avionics Systems Conference, vol. 2, pages 8.0.2.1 -
vs. time which is calculated during the path planning of the 8.0.2.12.
robot. The CPU usage is also noted during the path calculation [14]. Russell .S, P. Norvig [2003] "Artificial Intelligence: A Modern
Approach", Prentice Hall, pages 97-104.
to identifY the performance of the robot under different loads [15]. Stentz .A [May 1994] "Optimal and Efficient Path Planning for
and path fmdings. Partially-Known Environments", Proceedings of the IEEE
International Conference on Robotics and Automation, pages 3310
VII. CONCLUSION - 333

This project implemented a simulated annealing based

approach to deal with path planning issue for robots in
dynamic environment - this is one of the main contributions
of the project. The approach uses the vertices of the static and
dynamic obstacles as search space to obtain the optimal path