Reinforcement learning in mobile robot navigation : A literature review

Abstract

Reinforcement learning (RL) is a promising approach for enabling the autonomous control of mobile robots, especially in navigation tasks. Unlike classical model-based control methods, it enables agents to learn control policies by interacting with the environment, making it suited for changing and unpredictable settings. This thesis presents a structured literature review of reinforcement learning in mobile robot navigation. The core concepts of reinforcement learning and different RL methods are introduced to provide a basic understanding of the topic. Both classical and deep reinforcement learning (DRL) methods are presented. Different issues in reinforcement learning navigation are discussed in detail. Finally, conclusions are drawn about the current state and prospects of RL navigation.