Adaptive Learning Approach of Fuzzy Logic Controller with Evolution for Pursuit-Evasion Games

This paper studies a simplified pursuit-evasion problem. We assume that the evader moves with constant speed along a trajectory that is well-defined and known a priori. The objective of steering control of the pursuer modeled as a nonholonomic unicycle-type mobile robot is to intercept the moving evader. An adaptive learning approach of fuzzy logic controller is developed as an inverse kinematics solver of unicycle to enable a mobile robot to use the evader trajectory to adapt its control actions to pursuit-evasion game. In this proposed approach, GA evolves the parameter values of the fuzzy logic control system aiming to approximate the inverse kinematics of pursuer so as to generate a trajectory capturing the evader. Simulation results of pursuit-evasion game illustrate the performance of the proposed approach.