Research on AGV Path Planning Based on Reinforcement Learning

This article is based on reinforcement learning theory, constructs a complete real-time environment map, and designs a dynamic path planning method for Automated Guide Vehicle(Hereinafter referred to as AGV) in mixed flow workshops. In view of the problems of traditional reinforcement learning algorithm in path planning, this paper introduces the theory of prior knowledge, designs a heuristic reward function, proposes an action selection strategy with harmonic function, and constructs an improved reinforcement learning algorithm for AGV path planning, thus strengthening the effectiveness of AGV motion path exploration. Comparative simulation shows that the improved reinforcement learning algorithm converges faster, improves the reachability of environment map edges, and shortens the path finding time.