Physics-Informed Data-Based LPV Modeling and Validations of Lateral Vehicle Dynamics

This article introduces an integrated physics-data-based approach to derive a linear parameter varying (LPV) model for nonlinear lateral vehicle dynamics. The proposed method can accurately capture the nonlinearity of the lateral vehicle dynamics using only a few data points without model training or calibrations. Specifically, a novel nonlinear tire force representation based on time-varying stiffness is first proposed. Then, from the physical principles of vehicle motions, the nonlinear lateral dynamics are expressed in the LPV formulation, where the scheduling parameters are the longitudinal speed and slip ratio. After that, a data-based LPV representation is formulated based on a small set of data points satisfying the condition of persistent excitation. The system evolution of nonlinear lateral dynamics can be precisely recovered by parametric matrices formed by these data snapshots. Simulation data from
CarSim
and experimental data from multiple commercial vehicles are utilized to validate the proposed physics-informed data-based LPV model. The results demonstrate exceptional accuracy and noise-handling ability of the proposed modeling approach across multiple scenarios and vehicles.