Using Genetic Fuzzy Algorithm for Robot path planning

Most systems are nonlinear in nature and generally there aren't analytic control system design methods for them. Using artificial intelligent methods such as fuzzy logic is more effective for such systems. But generating fuzzy rules is the major problem of fuzzy logic controllers. Sometimes fuzzy rules are obtained from numerical data and occasionally rules are derived from human experts who have acquired their knowledge through experience. However these experiences may not always be available. Thus selection of an optimal or suboptimal set of rules from the set of all possible rules is an important and an essential step toward the design of any successful fuzzy logic controller. We used the genetic algorithm to generate fuzzy rules. Therefore we don't require any expert knowledge and input-output data. In this paper, the backing of a track to a loading dock is selected as a nonlinear system. The membership functions with different number of fuzzy sets are applied for the input and output linguistic variables. In fuzzy logic controller, generally one fuzzy rules set is applied during trajectory. Then we apply two approaches to arrive the final point with possible least steps. In first method, the input space is divided into two regions, the vicinity of the set point and the rest. We use fuzzy rules with less number of fuzzy subsets in away from the set point and we increase the number of fuzzy subsets of fuzzy rules set in the vicinity of the set point. Therefore we apply two fuzzy rules sets in two equal regions. In second method, we vary the set of fuzzy rules during trajectory based on least error. The performance of the first method is compared with the performance of second method and general method which one fuzzy rules set is applied during trajectory.