Pursuit-Evasion Game Based on Fuzzy Actor-Critic Learning with Obstacle in Continuous Environment

This paper employs the fuzzy actor-critic learning (FACL) and the Kalman filter (KF) to tackle the pursuit-evasion game (PEG) within a continuous environment, considering a scenario involving multiple pursuers and a single evader. We design reasonable reward functions for the pursuer and the evader, enabling them to complete the pursuit-evasion task and achieve obstacle avoidance. The strategies for both the pursuer and the evader are acquired through the FACL algorithm, while learning is extended from the discrete domain to the continuous domain. Additionally, pursuers use the KF to predict the evader's position, enhancing their ability to enclose and capture the evader. We demonstrate the advantage of the pursuers moving toward the evader using a geometric method, which compresses the evader's movement space and reduces capture time. The effectiveness of the proposed algorithm in capturing the evader and avoiding obstacles has been validated through simulation results.