Design of a genetic-fuzzy system for planning crab gaits of a six-legged robot

This paper describes a genetic-fuzzy system in which a genetic algorithm (GA) is used to improve the performance of a fuzzy logic controller (FLC). The proposed algorithm is tested on a number of gait-generation problems of a hexapod for crossing a ditch while moving on flat terrain along a straight line path with minimum number of legs on the ground and with maximum average kinematic margin of the ground-legs. Moreover, the hexapod will have to maintain its static stability while crossing the ditch. The movement of each leg of the hexapod is controlled by a separate fuzzy logic controller and a GA is used to find a set of good rules for each FLC from the author-defined large rule base. The optimized Cs are found to perform better than the author-designed FLCs. Although optimization is performed off-line, the hexapod can use these FLCs to navigate in real-world on-line scenarios. As an FLC is less expensive computationally, the computational complexity of the proposed algorithm will be less than that of the traditional methods of gait generation.