Continuous reinforcement learning algorithm for skills learning in an autonomous mobile robot

Abstract

Learning endows a mobile robot with a higher flexibility and allows it to adapt to changes occurring in the environment or in its internal state in order to improve its results. Based on this idea, this paper presents a reinforcement learning algorithm which allows the robot to learn simple skills such as go to goal and contour following. In the proposed learning algorithm the robot receives a real continuous reinforcement signal. Thus, it is not necessary to estimate an expected reward. Most of the robotic applications work with continuous variables such as velocity, position, sensors readings etc. The presented reinforcement learning algorithm is able to manage continuous input and output spaces. Finally, the robot is capable of performing the complex skill called go to go avoiding obstacles from the sequencing of previously learnt skills.