Data-driven Tuning of a NMPC Controller for a Virtual Motorcycle through Genetic Algorithm

In recent years, virtual prototyping tools have gained a central role in the design process of new vehicles. While vehicles dynamics emulation is nowadays extremely accurate, the reproduction, in a virtual environment, of a realistic driver behavior is still a challenging problem. For two-wheeled vehicles, the inherent instability of the motorcycle makes the design of a Virtual Rider (VR) a crucial element to ensure the effectiveness of the simulation result. Model Predictive Control (MPC) results to be a very promising technique to deal with this problem. However, a high number of parameters have to be calibrated, making the tuning procedure a major challenge. In this manuscript, a genetic algorithm for tuning a Nonlinear MPC-based VR is presented. The considered optimization objectives of the GA are both the performance and the robustness of the solution with respect to controller parameters variations. The proposed method results suitable for the tuning of the nonlinear MPC-based virtual motorcycle rider, obtaining better results with respect to previously published manual tuning. Moreover, the proposed robustness measure has proven to be consistent with the simulation results.