NAVIGATION FOR MOBILE ROBOT IN UNKNOWN ENVIRONMENT USING REINFORCEMENT LEARNING METHODS

Abstract

Mobile robots are used in a wide range of fields in the 4.0 industry. Using
mobile robots can provide such advantages as time efficiency, production
optimization, and so on. However, operating mobile robots with stability and
flexibility is always a big challenge and receives numerous researchers’ attention. In
particular, mobile robot navigation is considered a key technique in the robotics
field. Many research studies and technical methods have been proposed and
implemented to solve this issue, in which the reinforcement learning (RL) method
has attracted considerable attention because of its ability to learn from experience
and powerful adaptability. This paper proposes a navigation framework for mobile
robots based on a Q-Learning algorithm and obstacle avoidance method. The
relationship and characteristics between robot behaviors and environmental
conditions are also analyzed. The results of the Gazebo simulation are then
compared to those of SARSA to prove the efficiency of the proposed approach.