Estimating Modeling Errors in Vehicle Dynamics for Timely Detection of Tire Force Saturation

Safety-critical control of autonomous vehicles requires estimating modeling errors and uncertainties. This can augment the vehicle's ability to execute a safety-necessitated aggressive and evasive maneuver to avoid an accident while the tire forces are pushed toward saturation. This work experimentally validates an input-to-state stable nonlinear observer previously introduced by the authors to estimate errors in modeling dynamics of the bicycle model used for vehicle control. These errors are translated into improved estimates of tire-road cornering forces. Tracing the evolution of these forces against the lateral slip of the bicycle tires provides a qualitative measure of the vehicle's proximity to its limits-of-handling. The accuracy and robustness of the observer validated using cornering maneuvers in a one-tenth scale car demonstrates the potential of the observer for model correction and timely detection of tire-road force saturation.