Authors:
Abdelkrim Nemra
;
Hacene Rezine
and
Abdelkrim Souici
Affiliation:
Unit of Control, Robotic and Productic Laboratory, Polytechnical Military School, France
Keyword(s):
Reinforcement learning, Fuzzy controllers’ Genetic algorithm, mobile robotic.
Related
Ontology
Subjects/Areas/Topics:
Artificial Intelligence
;
Biomedical Engineering
;
Decision Support Systems
;
Expert Systems
;
Health Information Systems
;
Hybrid Learning Systems
;
Informatics in Control, Automation and Robotics
;
Intelligent Control Systems and Optimization
;
Knowledge Engineering and Ontology Development
;
Knowledge-Based Systems
;
Symbolic Systems
Abstract:
An efficient genetic reinforcement learning algorithm for designing Fuzzy Inference System (FIS) with out any priory knowledge is proposed in this paper. Reinforcement learning using Fuzzy Q-Learning (FQL) is applied to select the consequent action values of a fuzzy inference system, in this method, the consequent value is selected from a predefined value set which is kept unchanged during learning and if the optimal solution is not present in the randomly generated set, then the performance may be poor. Also genetic algorithms (Genetic Algorithm) are performed to on line search for better consequent and premises parameters based on the learned Q-values as adaptation function. In Fuzzy-Q-Learning Genetic Algorithm (FQLGA), memberships (premises) parameters are distributed equidistant and the consequent parts of fuzzy rules are randomly generated. The algorithm is validated in simulation and experimentation on mobile robot reactive navigation behaviors.