Adaptive neural net controller design

This paper presents a stability method which is based on the stability condition of sliding mode control to derive learning law for neural net controllers (NNC) to ensure the convergence of the training algorithm and the stability of the closed-loop system. The proposed method is an on-line approach of a multilayered neural network which does not required any information about the system dynamics and the lengthy training of the controller might be eliminated by using the proposed approach. The simulation results of a nonlinear system and a two-link manipulator demonstrate that the attractive features of the proposed approach include a smaller residual error and robustness against nonlinear interactions of an interconnected system or external disturbances.