Online Modeling of Lateral Vehicle Dynamics via Recursive Integrated Physics-Data-Based Method

This paper presents a recursive Integrated Physics-Data-Based (IPDB) algorithm to enable online modeling of lateral vehicle dynamics. The IPDB model integrates the fundamental physical laws and data snapshots in a moving-window framework, which possesses the properties of physical interpretations and adaptiveness simultaneously. Building upon our prior work on the IPDB approach, the recursive algorithm is derived to improve the computational efficiency. The recursive IPDB method updates the data-based system transition matrices using new data points rather than full-length data snapshots in the moving window. Besides, the proposed recursive IPDB method is proven that its modeling accuracy and data-based system representation are equivalent to the M-IPDB method. More importantly, the computational complexity is found to be independent of the moving window length, which allows handling larger window lengths with more data points to increase online modeling accuracy. Simulations by CarSim and experiments on commercial passenger vehicles are performed in various driving scenarios, where the recursive IPDB method is implemented and compared against the conventional IPDB approach. The results verify that the recursive IPDB method accurately captures nonlinear and time-varying characteristics of lateral dynamics, and produces equivalent modeling accuracy as the conventional IPDB method but with greatly reduced computation times