Nonlinear Autoregressive with Gaussian Process Regression-based Path-following Guidance for UAV under Time-varying Disturbances

Abstract

This paper investigates the problem of the degradation of trajectory tracking performance for an unmanned aerial vehicle under unknown external disturbances such as wind. We propose an adaptive guidance algorithm consisting of two parts: a baseline path-following guidance law with optimal error dynamics and a nonlinear autoregressive with Gaussian process regression (GP-NAR). Firstly, a baseline path following guidance law is constructed by adopting the guidance kinematics and the optimal error dynamics, assuming that the disturbances are measurable during this step. Then, the GP-NAR that the inputs are past observations of external disturbances is employed to capture and compensate for unknown external time-varying disturbances. Additionally, the partial autocorrelation function (PACF) is utilized to identify the lag value that determines the number of inputs of the GP-NAR model. Numerical simulation results demonstrate the performance of the proposed adaptive guidance algorithm under the unknown time-varying disturbances.