Sparse Gaussian Process Regression for Residual Dynamics Learning in Multi-rotor Aerial Vehicles Control

This paper introduces a method for modeling residual dynamics between a high-level planner and a low-level controller, using reference trajectory tracking in a cluttered environment as a case study. We aim to mitigate residual dynamics resulting solely from the kinematical modeling employed in high-level planning. Our high-level planner utilizes a simplified motion model for quadrotor motion. We propose a Sparse Gaussian Process Regression-based technique to model residual dynamics. In contrast, Data-Driven MPC, a recent technique, targets aggressive maneuvers without obstacle constraints. Our proposed method is compared with Data-Driven MPC in estimating residual dynamics error, including obstacle constraints. Comparative analysis indicates that our technique reduces nominal model error by an average factor of 2. Furthermore, we evaluate our complete framework against four other trajectory-tracking approaches in terms of tracking reference trajectory while avoiding collisions. Our approach demonstrates superior performance, achieving shorter flight times without sacrificing computational efficiency.