Abstract
State of the art vehicle dynamics control systems do not exploit tire road forces information, even though the vehicle behaviour is ultimately determined by the tire road interaction. Recent technological improvements allow to accurately measure and estimate these variables, making it possible to introduce such knowledge inside a control system. In this paper, a vehicle dynamics control architecture based on a direct longitudinal tire force feedback is proposed. The scheme is made by a nested architecture composed by an outer Model Predictive Control algorithm, written in spatial coordinates, and an inner longitudinal force feedback controller. The latter is composed by four classical Proportional-Integral controllers in anti-windup configuration, endowed with a suitably designed gain switching logic to cope with possible unfeasible references provided by the outer loop, avoiding instability. The proposed scheme is tested in simulation in a challenging scenario where the tracking of a spiral path on a slippery surface and the timing performance are handled simultaneously by the controller. The performance is compared with that of an inner slip-based controller, sharing the same outer Model Predictive Control loop. The results show comparable performance in presence of unfeasible force references, while higher robustness is achieved with respect to friction curve uncertainties.
Model Predictive Control With Learned Vehicle Dynamics for Autonomous Vehicle Path Tracking
