Improved Artificial Potential Field Based Ant Colony Optimization for Path Planning

Transcription

1 Available Online at International Journal of Computer Science and Mobile Computin A Monthly Journal of Computer Science and Information Technoloy ISSN 3 88X IMPACT ACTOR: 6.17 IJCSMC, Vol. 6, Issue. 7, July 17, p Improved Artificial Potential ield Based Ant Colony Optimization for Path Plannin Azal H.Nazzal Computer Science Department, University of Thi Qar, Iraq azal.alamery@mail.com Abstract- Path plannin field for mobile robot is an important problem that involves findin free path from initial position to oal position.in this paper we introduce an improved artificial potential field based on ant colony optimization.which can obtain additional ractive force to prevent artificial potential field from local minimum.and use ant colony optimization to determine the location of oal in the environment.we develop ant colony optimization to optimize planned path.we do not use the pheromone in all the path from initial position to oal position. We use pheromone in the oal position only. After that the proposed alorithm applied in a laptop computin equipped with 1.8GHz Intel core CPU,.5GB memory with 14 MATLAB in Windows 8.1 Operatin System. The simulation results confirm that the proposed alorithm could always find shortest path in less time. The comparison between ant colony optimization and enetic alorithm was also made in this paper. Keyword: path plannin, ant colony optimization, artificial potential field, motion plannin, multi robot. I. INTRODUCTION In the last few decades, robotic scientists have investiated on service mobile robots which could be able to operate within human-robot coexistent environments to execute different complex works, such as transportation of heavy objects, surveillance, rescue, and uidin people in exhibitions and museums. Autonomous mobile robot path plannin or naviation is one of the most important applications for robot control systems and has racted remarkable ention from number of researchers. We can classify the method for solvin the problem based on the environment into two types: known environment and unknown environment. And each of these two kinds could be divided into static and dynamic environment. In the known environment the robot know all the information about the environment before it start to move. the alorithms are suitable to solve the problem of path plannin for example artificial potential field [1],the Visibility Graph [],Voronoi diarams[3],the Grid method [4], the cell decomposition [5]. in unknown environment the robot does not have any information about the environment.in this state the robot must be solve the problem of path plannin by the sensor or the few information its available about the environment. A lot of researchers introduced important investiation results for examples enetic alorithm(ga)[6], ant colony optimization (ACO)[7], uzzy Loic [8], Neural Networks (NN)[9], particle swarm optimization(pso) [1].but these alorithms can t reach an optimal solution separately in complex environment. In this paper we introduce a new efficient approach for mobile robot motion plannin.we use two alorithms to solve the problem of path plannin: artificial potential field(ap) and ant colony optimization.ant colony optimization (ACO) is used to increase the 17, IJCSMC All Rihts Reserved 114

2 ractive force.the pheromone information enerated from ACO is used to prevent (AP) from fallin into local minimum. The remainder of this paper is structured as follows. In section displays ant colony optimization. Section 3 displays artificial potential field. In section 4 displays the proposed alorithm. Section 5 discusses the simulation and result for this paper and finally 6 the conclusion for this paper. II. ANT COLONY OPTIMIZATION Ant Colony Optimization is a meta-heuristic for solvin problems which need optimization. The idea of Ant colony Optimization was takin from the natural behaviour of ants in findin their short road from their livin place (nest) to food. The unique feature in the ant colony is that, it finds the shortest path from the nest to the food resource without any direct communication between the ants. This is the basic idea of swarm intellience that is noted not only in ants, it can be noted in birds, fishes, etc. The communication between the ants is based on the amount of the pheromone throuh their path. The difference in the amount of pheromone putted in the path taken by ants will shows the quality of the route selected to reach the food. [11] A. Improved ant colony optimization Oriinal ant colony optimization (ACO) contains some limitations, such as occupy lon time, ettin into local minimum if the size of the problem is lare. In oriinal ant colony (ACO), the amount of pheromone of all elements are same started. So all the solutions are started blindly in the beinnin of evolution phase, and it would occupy a lon time to find a short path from many paths. So we should improve the ACO. We don t need to put the pheromone in all the lenth of the path. We want to increase the ractive force to prevent the (AP) from local minimum. Therefore the robot moves in the environment to explore the entire environment because the robots don't have any information about the environment and determine the oal and obstacles locations by the pheromone puttin in the position of the oal and the obstacles only. Amount the pheromone differ between the oals and obstacles location. Amount of pheromone in the oals location is very much while in the obstacles locations is little. III. ARTIICIAL POTENTIAL IELD AP mainly contains from two force vectors, caused by the obstacles and oals positions. The oal position enerate the ractive forces to ract the robot to the oal while the obstacle position enerate ulsive force to el the robot from the obstacle.sum of these forces is controllin on the movement of the robot. [1] A. Attractive and ulsive force In artificial potential field, the oal position has ractive force on robots, and when the distance between the robots and oals increases, the ractive force increases. Position of robots in its work space referred to as q x y T. The ractive field created by the oal is iven by: U 1 q q kp oalq 1 1 Where k is a positive coefficient for AP, q x, y T is the location of oal. P q q q // oal // Is the distance from the location of robot to the position of oal. The ractive force is calculated by the neative radient of the ractive potential and iven the followin form: 1 q U q kp oalq kq q 17, IJCSMC All Rihts Reserved 115

3 q is a vector directed toward q with manitude linearly related to the distance from q to ulsive force has the form: 1 1 1, p q p U q p q p, pq, qob p Where is a positive scalin factor, let x 3, y q. The q be unique confiuration in obstacle closest to q. pq q q is the shortest distance between robot and obstacle. p is the larest impact distance of sinle obstacle. The ulsive force is calculated by the neative radient of the ulsive potential and iven the followin form: 1 1 q U q pq p p q 1 p q, p, p q q p p The total artificial force is calculated: q q is artificial force, 5 4 Where is ractive force and is ulsive force. When the environment has multi obstacles, the total ulsive potential field is the sum of all obstacles' ulsive potential field. The total potential field can be expressed as: n q q q 6 i1 Where n: is the number of obstacle IV. THE PROPOSED ALGORITHM This paper improved artificial potential field by usin the pheromone that enerated from ant colony optimization (ACO) as additional force to increase the ractive forces that ract the robot which the robot will not fall into local minimum.the work of ant colony optimization is explorin all the environment when it find the oal the robot put pheromone in the oal location (the amount of pheromone is very much )while puttin the little pheromone in obstacles location when it find obstacles. The pheromone amount in obstacle location is very little and does not affect at the total forces.the total ractive force that it racts the robot is: T 7 Where p T is the total ractive force, is ractive force that enerate from oal position, p is the pheromone force. Now each robot is drivin by the composition of three forces: T t Where 8 t the total force is drivin the robots, T is the total ractive force, ulsive force. The entire alorithm of our proposed method is as follow: 1. Select one robot randomly from multi robot.. The robot moves in the environment to explore the environment. 3. When it find oal, it puts the pheromone in the oal location. 4. When it finds obstacles, it puts the pheromone in the obstacles location but the amount of pheromone is very little. 5. Repeat until exploration the entire environment is finishin. 17, IJCSMC All Rihts Reserved 116

4 6. Compute the artificial potential force at current confiuration under our proposed improved artificial potential field. 7. Move all robots to the environment. 8. Each robot has ulsive force with other robot to avoid the collision. 9. Each robot reaches to the oal without any collision with obstacles. The robot is startin the movement from the first point of the environment to explore the environment accordin to the path planned by lobal planner.when it find oals in the path of robot it put some pheromone in the location of oals. The amount of pheromone in the location of oals is very much.when the robot finds the obstacle in the environment; it also put the pheromone in the location of obstacles. The amount of pheromone in the location of obstacle is little. In this paper are used multi robot, multi obstacle and one oal. The oal of the paper at first exploration the entire environment after that each robot must be reaches to the oal in the environment and avoids the collision with obstacles. V. SIMULATION AND THE RESULT Some simulation experiments are carried out for validin the proposed alorithm usin MATLAB. We used ant colony optimization to explore the environment by one robot. The environment is settin as square which shown in i.1. The paper has used different types of environments. Some environments were simple (number of obstacles is few and size of the environment is small). And some of them were very complex (number obstacles is too many and size of the environment is bi). When the robot find oal it puts the reen box in the location of oal and when it finds the obstacles it put the red box in the location of obstacles.i. shows that: i. 1 Simulation environment i. exploration the environment 15*15 17, IJCSMC All Rihts Reserved 117

5 After that the robots calculate the (AP) and start to move to the oal. The simulation results shown in the fiures, the reen box and red box resent the oal and the obstacles position respectively, the red line resents the path of the robot to reach to the oal.i.3 to i.5 display the path of robot to reach the oal in different environments. i. 3 displays path of robot in environment * i. 4 displays path of robot in environment 18*18 i. 5 displays path of robot in environment 5*5 Then we used three robots all of them are movin at the same time without collision between each other. Each robot must be reach to the oal in the environment. In i.6 black box resents the first robot, Purple box resents the second robot and Turquoise box resents the third robot. 17, IJCSMC All Rihts Reserved 118

6 i. 6 movement of multi robot TABLE I THE RESULTS O PROPOSED ALGORITHM Environment 18*18 No. obstacles :7 * No. obstacles:78 5*5 No. obstacles:8 Time (s) Distance (m) "Table I" resents the results for the proposed alorithm. We also compare the performance of the proposed alorithm and that of GA in [13].we use the same environment that they used with GA. rom "Table II", we can see that the performance of the proposed work better than the GA in the time and distance. TABLE II COMPARISON BETWEEN THE PROPOSED ALGORITHM AND GA Distance Time (s) (m) The 18* proposed No. obstacles alorithm :78 5* No.obstacles:8 15* No. obstacles :7 GA Iteration Iteration Iteration Iteration , IJCSMC All Rihts Reserved 119

7 VI. CONCLUSION Path plannin problem is one of the most important robotic problems for mobile robot to accomplish iven tasks. An improved artificial potential field ant colony optimization was proposed to obtain shortest path in less time without local minimum in unknown and static environment. Additional ractive force are utilized to eliminate local minimum caused by traditional AP when the composition of ractive force and ulsive force becomes zero. Exploration problem was solved by usin the pheromone to determine the oal and obstacles position. This alorithm has been successfully applied throuh the MATLAB proram without any problem. The paper tested the alorithm on a simple environments and complex environment. We use the environment with 3 robots, number of obstacles and one oal. The last work, the paper compares the performance of the proposed alorithm and that of GA. we see that the performance of the proposed work overmatches that of GA obviously as shown in "Table II". REERENCES [1] Q.Aao, Y.Huan and Zhou,AN evolutionary artificial potential field alorithm for dynamic path plannin of mobile robot, International conference proceedins on intellient robots and system, paes ,6. [] O. Aichholzer, M. Cetina, R. abila-monroy, J. Lea nos and J. Urrutia, Convexifyin monotone polyons while maintainin internal visibility, XIV Spanish Meetin on Computational Geometry,11. [3] C.O Dunlain and C.Yap., Are traction method for plannin the motion of a disc, journal of alorithms,vol. 6,paes , 198. [4] D.Payton, J.Rosenbl and D.Keirsey, Grid based mappin for autonomous mobile robot, Robotics and Autonomous systems, vol. 11, paes 13-1,1993. [5] C. H. Cai and S. errai,information-driven Sensor Path Plannin by Approximate Cell Decomposition, IEEE Trans on Systems, Man, and Cybernetics, Part B: Cybernetics, Vol.39, No.3, pp ,9. [6] Tamilselvi, Mercy shalinie, Hariharasudan,11, Optimal Path Selection for Mobile Robot Naviation Usin Genetic Alorithm,International Journal of Computer Science Issues, Vol. 8, Issue 4, No 1,11. [7]GUO Yu, LI Shi-yon.,Path Plannin for Robot Based on Improved Ant Colony Alorithm, Computer Measurement and Control, 17(1), pp ,9. [8] Mrs. Anita Gutta, Dr. Priti Srinivas Sajja,13, uzzy Loic based Intellient armin Multi Aent System,Anita Gutta et al,int.j.computer Technoloy & Applications,, Vol 4, paes8-3,13. [9] Youssef Bassil, 1, Neural Network Model for Path-Plannin Of Robotic Rover Systems, International Journal of Science and Technoloy, Vol., No.,1. [1] Micael S. Couceiro, Rui P. Rocha and Nuno M.. erreira, A Novel Multi-Robot Exploration Approach based on Particle Swarm Optimization Alorithms, IEEE International Symposium on Safety, Security and Rescue Robotics,11. [11] GUO Yu, LI Shi-yon,Path Plannin for Robot Based on Improved Ant Colony Alorithm, Computer Measurement and Control, 17(1):pp ,9. [1] Guanhui Li, Atsushi Yamashita and Hajime Asama, Yusuke Tamura, An Efficient Improved Artificial Potential ield Based Reression Search Method for Robot Path Plannin, IEEE International Conference on Mechatronics and Automation1. [13] Nadia Adnan Shiltah, Lana Dalawr Jalal, Path Plannin of Intellient Mobile Robot Usin Modified Genetic Alorithm, International Journal of Soft Computin and Enineerin, Volume-3, Issue-, , IJCSMC All Rihts Reserved 1