Racecar Longitudinal Control in Unknown and Highly-Varying Driving Conditions

This paper focuses on racecar longitudinal control with highly-varying driving conditions. The main factors affecting the dynamic behavior of a vehicle, including aerodynamic forces, wheel rolling resistance, traction force resulting from changing tire-road interaction as well as the occurrence of sudden wind gusts or the presence of persistent winds, are considered and assumed to have unknown models. By exploiting the theory on delayed input-state observers and using measurement data about the vehicle and wheel speeds, a dynamic filter that allows the online reconstruction of the above-mentioned unknown time-varying quantities is derived. Moreover, by exploiting the notion of effective tire radius, a reduced-degree-of-freedom model for the longitudinal vehicle dynamics is obtained, which is independent of the traction force and that enables, when used with the observer filter described above, an accurate speed control compensating for the resistance forces. One appealing feature of the proposed estimation and control method is that it requires no model information about such forces, for which, at the state-of-the-art, only heuristic approximations to be a-priori identified are available. Its effectiveness is shown via the simulation of scenarios where the car is required to execute aggressive maneuvers and the asphalt road surface abruptly changes from dry to wet, snowy, and icy. The evaluation also reveals that the proposed estimation technique outperforms standard solutions even in the presence of measurement noise.