Research on Path Planning of Autonomous Mobile Robot Based on Reinforcement Learning

Abstract

This paper presents a novel approach to optimize the path-planning capability of autonomous mobile robots, incorporating Q-learning (DQN). That is, deep Q-network (DQN) and the Prioritized Experience Replay (PER) strategy. DQN simulates the behavior of Quantum dots, it can show superior learning efficiency and generalization ability when dealing with a large number of state Spaces and actions. The introduction of PER mechanism reassesses the importance of experience samples in the learning process. By giving higher priority to those events that can cause large prediction error or have high reward, PER ensures that the algorithm can focus more on learning from the most critical experience. The reinforcement learning framework combined with DQN and PER can not only effectively avoid obstacles, but also dynamically adjust the path to deal with unexpected situations. In order to improve the robot navigation intelligence and decision-making speed in complex environments.