Model reference neural network control strategy for flight simulator

Abstract

A control scheme combining novel model reference adaptive control (MRAC) and neural network (NN) is proposed in this paper to achieve high tracking precision for servo systems. This scheme comprises an MRAC controller and an online NN controller in the velocity-loop and a traditional PD controller in the position-loop. For reducing influences which arose from modeling error, unknown model dynamics, parameter variation and disturbance acted on the velocity-loop, the NN controller is introduced to reduce the various influences mentioned above, and to adjust system to track the nominal velocity-loop reference model. Especially, as an innovation, a robust item is adopted to guarantee the system globally steady. Based on Lyapunov stability theory, updating algorithm of the weights of the NN controller, parameters of the MRAC and robust item are designed. Experiment results demonstrate that the proposed strategy can achieve high tracking precision for real-time position close-loop servo system.