An improved Q-Learning algorithm and its application to the optimized path planning for unmanned ground robot with obstacle avoidance

Abstract

In order to solve the problems in the optimum path planning of autonomous mobile robot in unknown complex environment, such as slow convergence rate and contact collision with obstacles, this paper proposes a modified Q-Learning algorithm based on flower pollination algorithm (modified Q-Learning with flower pollination algorithm, MQ-FPA). Firstly, Q-Learning is used to establish interactive relationship between autonomous mobile robot and environment; Then Flower pollination algorithm (FPA) is used to modify Q-Learning algorithm by initializing Q-table, so that the robot can explore the environment according to the transcendental knowledge gained according to FPA, which can improve the convergence rate of Q-Learning algorithm. The simulation results show that, compared with the traditional Q-Learning algorithm, the proposed MQ-FPA algorithm has the advantages of better adaptability, faster path optimization speed and better obstacle avoidance performance in different obstacle environment.