Professional Documents
Culture Documents
Logo Paper 09
Logo Paper 09
43-41 43
ISSN 2230-7621 © MIT Publications
ABSTRACT
Path planning of a robot [1] is a problem in which a problem space or a problem domain with a number of obstacles is given
and the aim is to find a collision free path which a robot can follow in order to reach its destination from the start position. Here
in this paper we are going to analyse various proposed techniques to find a path for a mobile robot in static and dynamic
environment. We have found a no. of techniques to find the collision free path with less effort and less cost.
1. INTRODUCTION
The moment we try to find out a solution for path planning of a
mobile robot[2], the first thing which came to our mind is about
some sort of sensors mounted on the head of robot that are sup-
posed to sense the obstacles and other difficulties that may oc-
cur in calculating the path of that particular robot. Secondly we
think about the nature of obstacles that may be static or dynamic
(moving obstacles). This paper is mainly devoted to analyze the
various approaches to find the ultimate solution of the above
said problem and locate the best one among them. One issue is
to produce a path that satisfies the criteria of optimality. Very
often the collision free path is highly desirable to be optimal or
near-to-optimal with respect to time, distance or energy.[3]
the part of their strings following the common number, if not forward but we can reduce efficiently the chromosome length to
another string is selected as the second parent and the same pro- reduce the required computation power. From the map topol-
cedure is followed. For example using the map in figure 2: ogy, we have observed that the optimal path to reach the goal
Parent 1 : 0- 1-5- 10-1 1-12-1 5 point is related to the number of obstacles, If the number of static
obstacles is equal m, the shortest path consists at most of (m + 2)
Parent 2: 0-3-6-4-5-10-9-15 points or (m + l) linear segments. This relation assumes that the
Each parent have a common point 10, the underlined parts of obstacle shape is not complicated and can be considered as mass
each string are exchanged, yielding: the two children are point. For example, if there is no any obstacle (m = 0), the short-
Child 1 : 0- 1-5-10-9-15 est path consists of a one linear segment from the start to goal
point; (number of points =2). In figure 1 the number of static
Child 2: 0-3-6-4-5-10-1 1-12-15
obstacles m=3; the shortest path consists of 3 point (start, via,
After cross over children are checked to determine whether each goal) which is less than 5 (m + 2). In figure 2, m =7; the shortest
string has repeated number. If so, the part of string between the path from points 0 to point 15 is (0-4-6-7-9-15) (6 points) which
repeated numbers is cut off. Some correction then required be- is less than 9(m + 2). Thus the following equation will be used to
cause it might be the case that the child is not admissible solu- define the chromosome length.
tion. This approach makes the algorithm more complex. The
C=m+2 (1)
important question is how to make the binary coding so possible
and easy to be used in a fixed length chromosome definition. Where m is the number of static obstacles.
One of the more challenging aspects in our proposed method is C is the number of genes in the chromosome.
encoding each string in a binary code with fixed length. A path For the given map in figure 2, the chromosome length = (7 + 2)×
of a mobile robot is encoded using binary numbers where each 4 = 36 binary bits. The initial population of chromosomes can
gene (via point) in a chromosome is encoded by 4 bits binary as be randomly generated such that each chromosome has a ran-
in the Table 1. dom m genes (via points), while the start and goal points are
fixed in the population. 13, 15 9, 12, 13, 14 B. Evaluation The
choice of a fitness function is usually very specific to the prob-
lem under condition. The evaluation function of a chromosome
measures the objective cost function.
The distance of a path indicated by the chromosome is used to
calculate its fitness. Since the fitness should increase as the dis-
tance decreases. Thus, the fitness function (F) of a feasible path
is evaluated as:
1
m 1 ; Feasible path
F d(Pi , Pi 1 )
i 1 (2)
0 ; Infeasible path
3.2 Evaluation role to diversify the solution population. Therefore, it isnot nec-
A path can be either feasible (collision free) or in feasible because essary that a solution is better after it is mutated.
intermediate nodes can fall on any of the grids. Theevaluation
should be able to distinguish feasible and infeasiblepaths and
tell the difference of path qualities within eithercategory. The
evaluation function is presented below:
N
Fcos t (di i C) (4)
i 1
4.3.2. Mutation
The mutation operator selects a node randomly and exchanges it
with some other node which is not present on current path. The
solution after mutation does not ensure to be the feasible solu-
tion but may help if the node falling on obstacle is changed.
Fig. 5. Path generated by initial chromosomes Chromosomes before Mutation
for problem given in Fig. 1
(0,0) (1,4) (1,9) (4,7) (5,6) (7,9)
The feasible path is a collision free path which does not have
any node falling on the obstacle and also no segment of path
Mutation Point
should intersect an obstacle. The length of chromosome can vary
from 2 to Total no of grids. Chromosomes after Mutation
(0,0) (1,4) (1,9) (6,3) (5,6) (7,9)
4.2 Evaluation/Finding Cost of Path
The cost finding method should be such that it should be ca- 4.3.3 Rectifying Node
pable of discriminating between the feasible and infeasible paths It is used to step out the node falling on an obstacle out of the
and define their qualities. The formula is given as: obstacle and to a finest grid around the obstacle.
Cost = di + K × p
Where “di is the sum of distances of center of all nodes forming
the path, K is a constant based on practical conditions, and P is
the penalty of intersecting an obstacle. This evaluation doesn’t
discard infeasible paths but penalize them with some additional
cost. They are not discarded because they might become opti-
mal solution after applying some optimizing operators that may
decrease their cost.
Before Applying Operator
4.3. Genetic Operators used in Proposed Approach
There are two basic operators Crossover and Mutation which
are used to generate off springs. In order to make the algorithm
effective, 4 new operators are also developed. These operators
make use available problem specific knowledge including knowl-
edge of the environment.
4.3.1 Crossover
The crossover operator randomly chooses a node from Parent 1 After Applying Operator
and the other node from Parent 2 so as to split them in two parts.
MIT International Journal of Computer Science & Information Technology, Vol. 4, No. 1, Jan. 2014, pp. 43-41 48
ISSN 2230-7621 © MIT Publications
Fig. 8. Result 2
MIT International Journal of Computer Science & Information Technology, Vol. 4, No. 1, Jan. 2014, pp. 43-41 49
ISSN 2230-7621 © MIT Publications
5. SIMULATION OF PATH PLANNING OF MOBILE 5.1 Path Planning Of Robot with Dynamic Obstacles
ROBOT IN DYNAMIC ENVIRONMENT First of all in order to place the dynamic obstacles[9] we need to
Genetic Algorithms are basically the search algorithms based on mark the starting and the ending position of the obstacles so as
the mechanics of natural selection and natural genetics. They to keep moving between the specified positions. A sample envi-
combine survival of the fittest among string structures with a ronment is shown in Fig.10. Some assumptions are also to be
structured yet randomized information exchange to form a search kept in mind that are:
algorithm with some of the innovative flair of human search. In The dynamic obstacles are point obstacles.
every generation, a new set of artificial creatures (chromosomes)
is created using random bits and pieces of the fittest of the old. They have certain velocity.
An occasional new part is tried for good measure which is known They move in a to and fro motion.
as mutation. Genetic algorithms despite being randomized are They don’t have the power to think so they will not wait for
not a simple random walk. They efficiently use historical infor- the robot.
mation to speculate on new search points which can be used to
get a solution with improved performance.
5.2 Evaluation
A path can be either feasible (collision free) or infeasible be-
cause intermediate nodes can fall on any of the grids. The evalu-
ation should be able to distinguish feasible and infeasible paths
and tell the difference of path qualities within either category.
The evaluation function is presented below:
FCost = di + K*P
Cost = Fcost + N*P
Where Fcost is the cost of path planning in static environment
di is the Euclidean distance from source to destination
K is the penalty cost in for static collision
N is the penalty for each collision, for ex. N=50
P is the number of pixels colliding with dynamic obstacles.
This evaluation gives penalty to infeasible paths, but still keeps
them in the population because they might become good fea-
sible solutions after certain genetic transformations. Besides, this
evaluation allows some overlap between finesses of feasible and
infeasible solutions because a very poor feasible path is not nec-
essarily better than a very good near-feasible path in the sense of
evolving solutions. It is beneficial to give more chance to some
good infeasible solutions that are easily to be evolved to good
solutions. To save computational time, some information obtained
by the evaluation needs to be recorded so that later on it can be
used by some specialized genetic operators as heuristic knowl-
edge without re-calculation. The information includes feasibil-
ity (feasible or infeasible, node-infeasible or line-infeasible),
number of infeasible nodes or line segments, and which
Fig. 9. Flow chart for path planning in sstatic environment
obstacle(s) a path intersects.
MIT International Journal of Computer Science & Information Technology, Vol. 4, No. 1, Jan. 2014, pp. 43-41 50
ISSN 2230-7621 © MIT Publications
• The manuscripts should be written in correct English and must convey facts of importance to the engineering community and
technocrats.
• The paper must be original, and not submitted to any other agency for publication.
• The manuscripts should be divided into sections, viz., Abstract, Introduction, Material and Methods, Results and Discussions.
• The paper should be limited to eight pages and should be typed in double column format. Longer paper will be considered, but
may be subject to delayed publication.
• The author(s) should give, name, affiliation, and e-mail on a cover page along with brief abstract and key words.
• All tables and figures should be distinctly clear. A brief caption at the bottom of each figure and title at the top of each table
should be provided.
• References to other publications must be made in standard style – (author’s name and year of publication) e.g. (McGrath,
1987). At the end of the manuscript, a reference list must be provided in the following format: name of author(s), title, journal/
Conference, (year), volume (Number), Pages.
• Manuscripts submitted should be accompanied with Authorization Letter for publication in favor of Editor-in-chief and brief
resume of the author on a separate sheet.
• The submitted manuscript may be subjected to substantial editing so that the paper conforms to our style.
• For more details, please visit www.mitpublications.org/Author-Instructions-mit-63.html
Secretary,
MIT International Journal of Computer Science & Information Technology (MITIJCSIT)
Moradabad Institute of Technology
Ram Ganga Vihar, Phase-2
Moradabad 244001, (U.P.), India
Tel: 0591-2452412, Fax: 0591-2452207
MIT Website: http://www.mitmoradabad.edu.in
Journal Website: http://www.mitpublications.org
Journal E-mail: cs@mitpublications.org