Robust neural adaptive stabilization of nonlinear composite systems with measurement noise

Based on dynamic neural networks, we investigate the problem of adaptive stabilizing nonlinear composite systems whose states are contaminated with external disturbances that act additively. Lyapunov stability theory is used to guarantee a uniform ultimate boundedness property for the system states and all other signals in the closed loop. In addition, the performance criteria of the mean-square performance are provided to quantify the control performance of proposed method. The controller derived is smooth, and no a priori knowledge of an upper bound on the "optimal" weights is required. The weights of neural networks are updating on-line, and the off-line training phase is not required. Simulations performed on nonlinear composite systems of electromotors illustrate and clarify the approach.