Reinforcement learning and Path Planning for the navigation of Mobile Robots

Abstract

, Wheeled mobile robots, which are present in several fields of activities nowadays, are machines equipped with perception, reasoning, and action capabilities to navigate autonomously and safely in their environments. This autonomous navigation skill requires a combination of hardware and software resources to perform basic tasks such as path planning, obstacle avoidance, and motion control with respect of the current navigation situation Reinforcement learning is one of the intelligent methods adopted to address the challenges of autonomous navigation in dynamic environments. It is a technique based on the interaction between an agent whose goal is to learn an action policy and its environment. In this thesis, this work focuses on integrating artificial intelligence, especially the Q-Learning algorithm, into the field of mobile robotics in order to enable robots to make intelligent decisions while on the move in environments with obstacles. The goal is to improve the autonomy and adaptability of robots by learning from experience, without the need for explicit programming for each task