Design and implementation of a high-performance, nonlinear MPC-based virtual motorcycle rider.

Virtual prototyping tools are nowadays widespread among automotive manufacturers and a successful usage of these tools requires a reliable human-like driver/rider. Specifically addressing motorcycles, the design of such controllers, called Virtual Riders (VR), is still a challenging task. In this paper, we first analyze the state of the art of the available control strategies, with specific reference to those based on Model Predictive Control (MPC), and then we describe a Nonlinear MPC-based VR for high-performance maneuvers. The internal dynamics is based on the sliding plane model, avoiding non-holonimic constraints, and the computed controls are the derivative of the actual bike commands (steering angle, throttle and brake effort). The NMPC problem is solved using an open-source tool (MATMPC) that is written in MATLAB and allows for easy development and tuning. The results are shown in cosimulation with a realistic simulation software (VI-BikeRealTime) on an hairpin turn, reaching the performance limits of the motorcycle. Moreover, computational burden is shown to be real-time compatible.