Mobile robot control using reinforcement learning

Abstract

Mobile robotics is a rapidly evolving field within artificial intelligence, offering promising solutions for autonomous operation in dynamic and unpredictable environments. This thesis investigates the use of deep reinforcement learning techniques to control a differential drive mobile robot, with a particular emphasis on accurate tracking of predefined trajectories. The study combines theoretical foundations in kinematics and localization with advanced learning algorithms such as Q-Learning and Deep Deterministic Policy Gradient (DDPG). The developed control policies are implemented and tested in realistic simulation environments using MATLAB and the Gazebo simulator via the MATLAB-ROS interface. The results demonstrate that reinforcement learning enables efficient robot trajectory tracking, confirming the potential of integrating artificial intelligence with physical modeling to design intelligent and autonomous robotic systems .