Exploring reinforcement learning techniques in the realm of mobile robotics

Abstract

Mobile robots are intelligent machines that can move and perform tasks in different environments. The key factor enabling the autonomy of mobile robots lies in the reliability, safety, and robustness of their navigation systems, without the need for human intervention. Achieving such a high level of autonomy has required extensive research and development efforts, encompassing both classical approaches and the latest advancements in artificial intelligence (AI) techniques. This review paper specifically focuses on the deep reinforcement learning (DRL) techniques employed for mobile robots. It provides a comprehensive look into the most significant DRL-based navigation and control algorithms for mobile robots. Sub-components of mobile robot navigation perception, mapping, localisation, and motion planning are well delineated under the lens of DRL and conventional methods. Furthermore, it also acknowledges the need for further research to address the challenges and limitations associated with deploying mobile robots in real-world applications.