Recurrent neural networks control of dynamic systems with unknown input hysteresis

This paper deals with the control of dynamic systems preceded by an unknown hysteresis, where the hysteresis is modeled by a differential equation. By exploiting properties of the differential equation, a recurrent neural network is developed to construct a hysteresis inverse, which can compensate the affection of the input hysteresis. By using a traditional PD controller, the whole system tracks a desired trajectory within a specified precision. Simulation results verified the proposed schemes.