Deep Reinforcement Learning Based AGV Self-navigation Obstacle Avoidance Method

A local path optimization model and obstacle avoidance strategy based on Actor-Critic algorithm is proposed for the local obstacle avoidance problem of automatic guided vehicles in a complex workshop environment. In the complex working environment of the production workshop, we analyze the automatic obstacle avoidance problem of AGV trolley, establish the front and both sides of the AGV tentacle model and Markov decision process, and describe the local obstacle avoidance path in the form of virtual tentacles. And based on deep reinforcement learning to solve the path obstacle avoidance strategy, it is applied to the AGV self-navigation system. The dynamic obstacle avoidance performance of AGV is tested through simulation experiments, and the effectiveness of the proposed algorithm is verified by completing local obstacle avoidance path planning under global path guidance.