Application of the GA-PSO with the Fuzzy controller to the robot soccer
Department of Electrical Engineering, Southern Taiwan University, Tainan, R.O.C Juing-Shian Chiou, Ming-Yuan Shieh Shih-Wen Cheng, Kuo-Yang Wang, Yu-Chia Hu Outline Abstract Introduction Motion Fuzzy Controller Structure
GA-PSO Fuzzy Controller Design Method Simulation Results Conclusions Abstract In this paper we proposed the method of GA-PSO to adjust the rule of the fuzzy system with robot soccer. The experimental scenarios involved a five-versus-five soccer simulation and the MATLAB simulation. Introduction(1/2) We use the robot system as our test platform since it can be used to fully implement a multi-agent system. In this paper, we choose five-versus-five simulation platform. (Figure1) Figure 1. The Five-versus-Five simulation platform Introduction(2/2) To achieve an optimal design for a soccer robot, we use Genetic algorithms- particle swarm optimization (GA-PSO) to adjustment fuzzy membership function thus reaches the optimal result. Motion Fuzzy Controller Structure(1/7)
In this part, we start design the fuzzy logic controller aimed at producing the velocities of the robot right and left wheel. We set two input parameters of the fuzzy logic controller are distance and angle . The former is the distance between the robot and the goal. The latteris the direction of with on the straight line path to the goal. Both are shown in Figure 2. Figure 2. the relation ofand Motion Fuzzy Controller Structure(2/7)
We set the values of variable , ,,,,,, and design two fuzzy controllers to control the velocity of the right and left wheels to move the robot. The fuzzy rules on which were based these fuzzy controllers are described in tables 1 and 2, and can be described according to the following equations: Then (1) (2) Motion Fuzzy Controller Structure(3/7)
Table 1. Fuzzy rule base of the left- wheel velocity fuzzy controller Table 2. Fuzzy rule base of the right- wheel velocity controller Motion Fuzzy Controller Structure(4/7)
The following term sets were used to describe the fuzzy sets of each input and output fuzzy variables: (3) (4) Motion Fuzzy Controller Structure(5/7)
As show in figure 3, the triangle membership function and the singleton membership function are used to describe the fuzzy sets of input variables and output variables. (a) (b) Figure 3. Membership function:(a) the fuzzy sets for; (b) the fuzzy sets for Motion Fuzzy Controller Structure(6/7)
Based on the weighted average method, the final output of these fuzzy controllers can be described by means of equation (5) and (6) Where and were determined according to Equations (7) and (8). (5) (6) (7) (8) Motion Fuzzy Controller Structure(7/7)
When the input data of andare given,and can be determined by using Equations (5) and (6) Thus, the left-wheel velocity and the right-wheel velocity can be obtained. GA-PSO Fuzzy Controller Design Method(1/11)
The procedure of GA-PSO algorithm can be described as follows: Step 1: Initialize the PSO algorithm by setting , the maximum number of generation (G), the number of particles (N), and four parameter values of and GA-PSO Fuzzy Controller Design Method(2/11)
Step 2: Generate the initial position vector and the initial velocity vector of N particles randomly by (12) and (13) GA-PSO Fuzzy Controller Design Method(3/11)
Step 3: Calculate the fitness value of each particle in the g-th generation by Step 4: Determine and for each particle by (15) and (16) GA-PSO Fuzzy Controller Design Method(4/11)
Step 5: Find an index q of the particle with the highest fitness by (17) and determine and by (18) and (19) GA-PSO Fuzzy Controller Design Method(5/11)
Step 6: If g=G, then go to Step 12, Otherwise, go to Step 7. Step 7: Update the velocity vector of each particle by (20) is a weight value and defined by (21) GA-PSO Fuzzy Controller Design Method(6/11)
Step 8: With fixed-length chromosomes that the problem is variable domains, select the number of chromosome population is , the crossover rate is , the mutation rate is Step 9: The definition of adaptive function to measure the problem domain on a single chromosome of the performance or adaptability. Adaptive function is built on the reproductive process, the basis for selecting pairs of chromosomes. Step 10: The size of a randomly generated initial population of chromosomes GA-PSO Fuzzy Controller Design Method(7/11)
Step 11: Calculating the adaptability of each chromosomes. Step 12: In the current population, select a pair of chromosomes Parental chromosomes are selected and their adaptability related to the rate. Adaptive chromosomes are selected with high rate is higher than the low adaptability of the chromosomes. Step 13: Through the implementation of genetic operators-crossover and mutation of a pair of offspring chromosomes. GA-PSO Fuzzy Controller Design Method(8/11)
Step 14: The offspring chromosomes into new populations. Step 15: Repeat step 13, unit the new chromosome population size is equal to the size of initial population Step 16: With the new (offspring) chromosome populations to replace to the initial (parent) chromosome populations. Step 17: Back to step 12, repeat this process until you meet the conditions for ending to stop. GA-PSO Fuzzy Controller Design Method(9/11)
Step 18: Check the velocity constraint by (22) Step 19: Update the position vector of each particle by (23) GA-PSO Fuzzy Controller Design Method(10/11)
Step 20: Bound the updated position vector of each particle in the searching range by (24) Step 21: Let g=g+1 and go to Step 3. GA-PSO Fuzzy Controller Design Method(11/11)
Step 22: Determine the corresponding fuzzy controller based on the position of the particle with the best fitness value In the above reasoning, we will use the FIRA simulator to confirm the results of our reasoning. Simulation Results(1/2)
The membership functions of and, as determined by the proposed GA-PSO based method, are presented in figure 4. (a)(b)(c) Figure 4. membership functions of (a) (b) (c) , as determined by the proposed GA-PSO method Simulation Results(2/2)
The figure 5 are the orbit of the soccer robot when controller by the proposed GA-PSO method and simulation with a FIRA simulation. Figure5. The soccer robot moving Conclusions The final results showed that, although the GA-PSO's convergence time is not the time than the PSO-based fast, and as we join the GA algorithm , after the results obtained will be closer to its optimal solution. In the future, we need to explore ways to let a faster convergence time for change Thanks for your attention !