Predictor-based neural network control for unmanned aerial vehicles with input quantization: design and application

Abstract

In this article, I design a predictor-based neural network (NN) controller for unmanned aerial vehicles (UAVs) with input quantization to address the trajectory tracking problem in the presence of time-varying disturbances caused by aerodynamics and external environment. The NN with a state predictor (SP) is employed in the controller design to improve transient performance without high-frequency oscillations and address the problem of instability caused by the time-varying disturbances. Additionally, the prediction errors from the SP are used to update the learning rate of the NN, resulting in smoother and faster learning responses. Furthermore, a hysteresis quantizer is employed to discretize signals and reduce the transmission burden on digital hardware, which can enhance the suitability of the system for practical implementation. Based on the Lyapunov method, the closed-loop system of the UAV achieves input-to-state stability (ISS). Finally, to validate and assess the performance and effectiveness of our proposed control method, I present and analyze both simulation results and experimental results from real-world applications.