Neural Network Based Adaptive Flight Control Using Feedback Error Learning

Abstract

[Abstract] This paper considers a real time adaptive flight control system using a neural network (NN). In this system, the NN aims to obtain the inverse dynamics of an aircraft, and the feedback error learning (FEL) strategy proposed by Kawato et al. is used as its learning scheme. In FEL, a NN is placed parallel to a conventional feedback (CFB) controller and works as a feedforward controller to complement the time delay in the CFB controller. Moreover, the adaptability of the NN enables the entire control system to exhibit a good performance under the fluctuation of aircraft dynamic characteristics. Numerical simulations and flight tests were carried out by imitating aileron actuator failure and fluctuation of aircraft dynamics. These results show that the proposed control system is able to improve the entire control performance.