Identification and control of dynamical systems using modified neural networks

The feedforward neural network, whose inputs are modified by the dynamical error feedback and the network derivatives, was proposed to reduce the inherent network approximation error. The identification model constructed by the proposed network and stable filters was obtained, which is used for the continuous time nonlinear system to improve the modeling performance. The adaptive control scheme for the nonlinear system based on the proposed network was developed. The simulation results have proved that the model could effectively be applied to control for a class of the nonlinear system.