Abstract: In this paper, an optimization algorithm, which is based on the self-organizing mapping of Kohonen's model, is proposed and applied to solve the problem that the fuzzy control rules of quadruped walking robot and other complex multivariable systems are too many. The fuzzy relation is identified from original fuzzy rules, furthermore, it is clustered by using Kohonen's network in order to optimize the fuzzy rules. The effectiveness of the algorithm is demonstrated by the simulation results.