Implementation of hybrid adaptive fuzzy sliding model control and evolutionary neural observer for biped robot systems

Abstract

In this paper, a hybrid adaptive fuzzy sliding-mode dynamic evolutionary neural network (DENN) technique is proposed for biped robot control. The hybrid control system, containing a key and a secondary controller, is implemented. The key controller combining an adaptive fuzzy sliding mode estimator and a DENN observer plays the role of the principal controller meanwhile the secondary controller is a compensator for the approximated error of the biped system uncertainty. The adaptive fuzzy controller is designed with two update rules to train the weights and the approximated fuzzy error of the estimator. Using Lyapunov principle, the adaptive fuzzy rules of the control system are derived so that the asymptotical stability of the biped system can be ensured. Eventually, simulation and experiment results confirm the efficiency, the robustness and the trajectory following performance of the new adaptive fuzzy control method for biped systems.