Reinforcement Learning for Mobile Robot's Environment Exploration

The field of mobile robotics has become increasingly significant in today's technology-driven world, with applications ranging from autonomous vehicles to automated delivery systems. This project aims to enhance the capabilities of mobile robots in navigating and exploring unfamiliar environments by developing and implementing a Reinforcement Learning (RL) model. The core objective is to enable these robots to autonomously explore, navigate, and adapt to new surroundings efficiently. By integrating the RL model into a mobile robot using the Robot Operating System (ROS), the project evaluates the system's efficiency through comprehensive ground experiments. The research compares the RL-based approach with established exploration methods, emphasizing the effectiveness of the Value Iteration algorithm, a dynamic programming algorithm for Markov Decision Processes (MDPs). Controlled experiments conducted in a 6×3 meter flat space with static obstacles and a 2D lidar sensor provide valuable insights into the RL model's performance and potential real-world applications. The findings highlight the advantages of the Value Iteration algorithm over traditional exploration methods, demonstrating its potential for improving autonomous robotic navigation and exploration.