Fuzzy Reinforcement Learning Rectilinear Follow-up of Trajectory per Robot

Youcef Dahmani¹ and Abdelkader Benyettou ²

¹ Laboratoire SIMPA, Ibn Khaldoun University Tiaret, Algeria

² Department of Computer Sciences, University of Sciences and Technology of Oran, Algeria

Abstract: Knowing the action space of an order, the objective consists in distributing this space in a certain set of actions equitably in order to choose the famous action among the candidate ones. This process is ensured by reinforcement learning aided by fuzzy logic. We have established an algorithm applying the fuzzy Q-learning with a fuzzy limited lexicon. We have applied it to a robot for the training of the follow-up of a rectilinear trajectory from a starting point “D” at an unspecified arrival point "A", while avoiding with the robot butting against a possible obstacle. The goal of this work tries to answer the question, in what the reinforcement learning applied to fuzzy logic can be of interest in the field of the reactive navigation of a mobile robot.

Keywords: Mobile robot, navigation, reinforcement learning, fuzzy logic, fuzzy Q-learning, fuzzy inference system.

Received April 21, 2004; accepted August 21, 2004