Chih-Hsin Tsai, Jing-Sin Liu, Kuo-Bin Tseng and Wei-Song Lin
In this paper, a learnable neuro-fuzzy controller is proposed for on-line implementing a decoupling control action for uncertain composite affine nonlinear plants to track a prescribed trajectory. In structure, the controller is composed of decentralized fuzzy systems with embedded two-stages rule credit assignments mechanism cascaded with an interconnections compensating associative memory network and a nonsingularity supervisor. In analytical form, the controller can be parametrized by a set of linear parameters, which represent a combination of of the credits of rules, locations and shape factors of membership functions. The parameters are tuned by a deadzone adaptation algorithm to compensate for uncertainties. It is shown that the incorporation of deadzone in controller guarantees the stability of adaptation in the neuro-fuzzy system and moreover, a given level of attenuation for tracking error in the presence of unknown but bounded interconnections and disturbances. Simulation results of SISO plant, an inverted pendulum, and MIMO plant, a two-link planar robot manipulator, are given to demonstrate the effectiveness and robustness of the neuro-fuzzy controller for output tracking in composite nonlinear systems.