A Q-Learning Strategy for Path Planning of Robots in Unknown Terrains

The goal of the mobile robot path planning is to find a path from start position to final position. Path should be collision free and the path should also be optimized for the shortest path. This paper is rooted on implementation of Q-Learning algorithm to make robot learn how to avoid unknown obstacles. When the robot learned to avoid obstacles, it used a hybrid controller that uses the learned Q-table to navigate in an unknown environment to reach the goal. The proposed algorithm used ROS2 (Robot Operating System 2) foxy and Gazebo 11 to simulate the proposed algorithm on a mobile robot. Further, results are depicted through simulation and experiments to exhibit the effects of introduced path planning algorithm.