Gaussian Process-based Adaptive Path-Following Guidance for Unmanned Aerial Vehicles

Abstract

In this paper, an adaptive path-following guidance algorithm is proposed by leveraging the Gaussian process regression (GPR) model. To this end, a baseline path-following guidance law is first derived by utilizing the guidance kinematics and the specific form of the error dynamics called the optimal error dynamics, under the assumption that unknown external disturbance is measurable. A GPR model is designed with the purpose of estimating the unknown disturbance term. The GPR model is then augmented to the baseline path-following guidance law by replacing the unknown disturbance term in the guidance command ith the GPR model output. To construct a dataset for training the GPR model, a nonlinear disturbance observer is used to determine the output variable of the GPR model. Finally, the proposed algorithm is tested through numerical simulations.