Neuro-fuzzy motion controller design using improved simple genetic algorithm

In this paper a method for the design of a neuro-fuzzy motion controller is proposed. The controller is aimed to ensure the continuous locomotion of a brachiation (double pendulum like) mobile robot (BMR). The design method consists of two stages. In the first stage the structure of a Takagi-Sugeno (TS) type of fuzzy controller (FC) is established. In the second stage the parameters of a feedforward type neural network structure that embeds the controller are derived. During the first stage, the parameters of both the antecedent and the consequent part of the fuzzy rules are obtained in the same time, using an improved simple genetic algorithm. The paper presents experimental results obtained using the simulated evolution of the BMR under the control of the designed controller. The advantages of the proposed method and the possibilities of further improvements are discussed.