Data-Driven MPC for Uncertain High-Order Fully Actuated Systems

This study combines machine learning with Model Predictive Control (MPC) to regulate the High-Order Fully Actuated (HOFA) nonlinear systems with disturbances. By exploiting the full-actuation characteristic of the HOFA system, we neatly eliminate the nonlinearity of the system, thus circumventing the complex computation of uncertainty propagation in nonlinear stochastic MPC. To address the disturbances, Principal Component Analysis (PCA) precisely identifies principal components, while Kernel Density Estimation (KDE) assesses probability distributions and uncertainty bounds. Furthermore, we apply stricter state constraints and define terminal sets to ensure the recursive feasibility of the method theoretically. Lastly, a numerical example is simulated to confirm the effectiveness and reliability of the proposed method.