Adaptive Neural Network Control for Nonlinear Systems with Unknown Hysteresis via H\infty Approaches

In this paper, in order to design control scheme to mitigate the effects of unknown hysteresis, a class of novel hysteresis models are proposed. We superpose a finite of many different deadband width backlash models, which represented as a dynamics to mimic hysteresis in actuator. With the model proposed, a single hidden layer neural network (NN-based) adaptive control scheme for nonlinear systems with unknown hysteresis nonlinearity is developed. The control scheme adopts the design method of pseudo-control. For the nonlinear dynamic systems, with time-varying external disturbance and strong nonlinearity and large uncertainty of unknown hysteresis, which output is not available, we adopt Hinfin optimal control techniques. Our result indicates that arbitrarily small attenuation level can be achieved via the proposed adaptive neural networks control algorithm if a weighting factor of control variable is adequately chosen. The effectiveness of the proposed control scheme is illustrated through simulation