A Learning-Based Nonlinear Model Predictive Controller for a Real Go-Kart Based on Black-Box Dynamics Modeling Through Gaussian Processes

Lately, nonlinear model predictive control (NMPC) has been successfully applied to (semi-) autonomous driving problems and proven be a very promising technique. However, accurate models for real vehicles could require costly time-demanding specific measurements. To address this problem, the exploitation of system data complement or derive prediction NMPC explored, employing learning dynamics approaches within learning-based (LbNMPC). Its application automotive field focused on discrete gray-box modeling, in which nominal dynamics is enhanced by data-driven component. In manuscript, we present an LbNMPC controller go-kart based continuous black-box accelerations obtained Gaussian processes (GP). We show effectiveness proposed approach testing vehicle, highlighting approximation steps required get exploitable GP real-time application.