Você está na página 1de 8

Shortest Path Problem Elucidation based on Particle Swarm

Optimization (PSO)
Rishu Mishra, Saifuddin Tariwala
mishrarishu17@gmail.com, saiftari@gmail.com
Department of Computer Science
CS652 Emergent and Adaptive Computation

Abstract
This paper proposes hybrid algorithm based on a Particle swarm optimization
(PSO) to solve the multiple destination routing (MDR) problems and single-
source shortest-path problem (SPP) commonly encountered in graph theory.
Shortest path problems occupy an important position in operations research as well
as in Artificial Intelligence. The single-source shortest path problem is defined as a
weighted directed graph ((V, E), w) and a source vertex s in V, find a shortest path
from s to every vertex of V. If a path has negative weight edges in a cycle, then a
shortest path is not defined. The problem has been proven to be NP-complete and
the traditional heuristics (e.g., the SPH, DNH and ADH) are inefficient in solving
it. The particle swarm optimization (PSO) is an efficient global search algorithm
and is promising in dealing with complex problems. A new encoding/decoding
scheme based on heuristics has been devised for representing the SPP parameters
as a particle in PSO. We will simulate several networks with random topologies to
illustrate the efficiency of the proposed hybrid algorithm for shortest-path
computation. The proposed algorithm can be used as a platform for solving other
NP-hard SPPs.

We will apply the above hybrid algorithm to control a robotic swarm to
collectively solve exploration and navigation tasks. We want to apply swarm
intelligence which takes inspiration from social insect colonies, to allow the swarm
of robots to solve complex tasks, while using simple control strategies for an
individual robot.

KEYWORDS

Particle Swarm Optimization (PSO), Single Source Shortest Path Problem (SSP)




INTRODUCTION

Particle swarm optimization (PSO) is a computational process which inspired by
the social behavior among the homogenous agents also known as particles in the
swarm space. The theory of Particle swarm optimization was given by Kennedy,
Eberhart and Shi. The goal of the algorithm is to find the optimal global and
personal location for every individual particle in the swarm space so that they can
result in the ultimate best global position of the entire swarm. PSO is a
metaheuristic process in which the particle is assigned some random space and
some initial random velocity. With every iteration, each particle changes position
with respect to their velocity and ultimately advancing to the position which is
optimal for the swarm space and to the particle itself.
After each iteration, the velocity of individual particle is calculated by the
following formula:-
Vj(T+1) = Vj(T) + (C1* rand() * (Pjbest Pj(T))) + (C2* rand()* (Pgbest-Pj(T))).

Where Vj(t+1) is the recent velocity of jth particle,C1 is the cost of personal best
position of the particle and C2 is the cost of the global best position of the particle
respectively, Pj(T) is the position of the jth particle at time T, the best known
personal position of the particle is represented by Pjbest and its best position
known to swarm is represented by Pgbest.
After having the new velocity, now we can have the new position by the following
formula:-
Pj(T+1)=pi(T) + Vj(T).

SHORTEST PATH PROBLEM:-

Shortest path problem in a graph theory is defined as finding the optimal path from
one node to another such the cost of reaching between two nodes is minimized.
Consider the graph G having set of vertices V and edges E, the cost between two
vertices Vi and Vj is represented by Cij. The objective is two find the path which
has the minimum cost.
There are many applications of SSP like the technique of finding direction between
two locations which is used by the websites like google, TSP is also based on
shortest path finding technique, in video games, these algorithms are frequently
used to find the shortest path between two points on a map. Path finding, as it is
called in this context, can be used by AI to plot routes, or by the game engine to
assist users in plotting routes. n the military we have certain small unmanned aerial
vehicles/devices which has some pre set point it needs to scout, and since it must
travel really far distances, on a small device like that it would be too expensive to
try to control it via satellite, and radio control would deteriorate before reaching
the farthest point. That is where the algorithm comes into play.

3. Design

Navigation is a major challenge for autonomous robotic swarm. (finding a collision
free motion for the robotic swarm from one configuration or state to another
known as Particle swarm optimization (PSO)).The most significant challenges
confront in autonomous robotics lie in the area of automatic path planning. Path
Planning has become a major research in robotic swarm.
Today, while obstacle avoidance remains a vital issue, other important constraints
are required to be considered as well (optimality, coverage, equilibrium etc.).

The Problem:
The problem is to find collision free path for the robotic swarm from the
start configuration to a goal configuration

Proposed Solution:
The Proposed algorithm computes an optimal path in order to arrive at the
goal position on the grid Particle swarm optimization (PSO).
Principle involved detection of a junction that implies the successive
increment in the coordinate system.
The robotic agent can arrive on the points that are situated in diagonal
position.

The Algorithm:
1. Receive coordinates and store them in an array variable
2. Compute & send checksum to the computer
3. Define a grid matrix m[x][y]=9.
4. Identify the trace point and obstacle point in the matrix and set :
trace point grid matrix value to 1
obstacle point grid matrix value to 8.
Grid matrix values:
neutral points: 9
trace points: 1
obstacle points: 8
5. Define Grid point value (GPV) Matrix ,v[x][y]


6. For h=0 to (no. of trace points) do
Set loop=0
while (loop!=1) do

7. Define possible future coordinates:

8. Define distance matrices:
ed=v[tx[h]][ty[h]]-v[ex][ey]
nd=v[tx[h]][ty[h]]-v[nx][ny]
wd=v[tx[h]][ty[h]]-v[wx][wy]
sd=v[tx[h]][ty[h]]-v[sx][sy]
ed is the difference in the next trace point GPV matrix values and the east
direction point GPV matrix value.








9. Check conditions for ed, nd, wd and sd for optimal path




end while
end for

10. Store all points whose Grid matrix value corresponds to 7 in a path array, path.


Results:
The proposed algorithm work successfully and effectively determines the optimal
path.
Optimal Path followed by robot:
{0x10 0x20 0x21 0x31 0x32 0x33 0x23 0x13 0x14 0x15 0x25 0x35 0x45 0x55}.

Green Dots: Trace points (2, 1) (3, 3) (1, 5)
Red Dots: Obstacle points (1, 2) (5, 3) (3, 4)
Blue Dot: Final Destination (5, 5)
Arrows: Motion of the Robot
White Line: Optimal path traversed on the Grid

Conclusions & Future Work:
The proposed algorithm have been successfully implemented to create an effective
solution in the areas of automatic offline path planning, exact grid navigation get
Particle swarm optimization (PSO).













References
1. Ammar W. Mohemmed, Nirod Chandra Sahoo, A New Particle Swarm Optimization
based Algorithm for Solving Shortest Path Tree problem. Proceedings of Press,.
Telecommunications and Malaysia International Conference on Communications 2007
ICTMICC 2007 IEEE International Conference on (2003).
2. Ammar W. Mohemmed, N C Sahoo, Tan Kim Geok, Particle swarm optimization with
noising metaheuristics for solving network shortest path problem. Telecommunications
and Malaysia International Conference on Communications 2007 ICTMICC 2007 IEEE
International Conference on (2007).
3. Leung, Y., Li, G., Xu, Z.B., A genetic algorithm for the multiple destination routing
problems. J. IEEE Trans. Evol. Comput. 2, 150-161 (1998).
4. Zhi-hui Zhan and Jun Zhang, Discrete Particle Swarm Optimization for Multiple
Destination Routing Problems, IEEE Transactions on Systems (2009).
5. Marc Szymanski, Tobias Breitling, Jorg Seyfried, Heinz Wrn, Distributed shortest-path
finding by a micro-robot swarm, Proceedings of the 5th international conference on Ant
Colony Optimization and Swarm Intelligence (2006).

Você também pode gostar