Abstract
A fuzzy logic controller (FLC), intended to plan and control the motion of a car-like robot during its navigation among several moving obstacles, is designed automatically using a genetic algorithm (GA). The FLC is expressed by utilizing the structure of a neural network (NN), and a GA is used to optimize both its data base as well as rule base. An importance factor is introduced to determine the strength of a rule and thus justify its redundancy, if any. Results of the proposed approach are compared to those of a neuro-fuzzy approach, genetic-neuro-fuzzy approach and a potential field method, through computer simulations, for solving the navigation problems of a car-like robot. The present approach is found to perform better than the other three approaches, for most randomly-generated test scenarios.
Get full access to this article
View all access options for this article.