Autonomous robot navigation based on machine mapping and learning techniques

This article presents a method of autonomous navigation for mobile robots using a hybrid architecture, composed of mapping and probabilistic techniques of reinforcement learning (RL). The robot must first learn the limits of the environment and how to move intelligently between distinct points. For the simulation environment we used the software Player and Stage, which made it possible to verify the behavior of the mobile robot used across the map. The mapping method used for the representation of the environment was based on grade of occupation, then was used to define the environment in the process of reinforcement learning. The both learning techniques Q-Learning and R-Learning have been implemented and compared. The methods demonstrated the ability of learning by the robot in order to successfully accomplish the goals of this work.