This paper presents an Integrated physics-data-based(IPDB)modeling of lateral vehicle dynamics with moving-window data snapshots.The IPDB model encodes the fundamental physical principle of four-wheel vehicle motions ...This paper presents an Integrated physics-data-based(IPDB)modeling of lateral vehicle dynamics with moving-window data snapshots.The IPDB model encodes the fundamental physical principle of four-wheel vehicle motions and simultaneously carries out the adaptiveness of the data-driven approach.Specifically,the traditional physics-based lateral dynamics considering four-wheel interaction are first derived into an affine linear-parameter-varying model,in which the vehicle-related parameters and motion variables are separated.Then,by using the Kronecker product,the IPDB model,directly formulated by the data snapshots in the moving-window fashion,is obtained for system representation.As a result,the IPDB technique rendered model is physically interpretable.The impacts of moving window length on modeling performance are numerically studied.The IPDB model accuracy is validated with data from CarSim simulations and experiments with passenger vehicles under various scenarios.It is further demonstrated that the proposed IPDB model is more data-efficient than other data-driven methods since it only uses the data snapshot in the moving window to update the model recursively.This characteristic enables the IPDB method with online modeling capability to adapt to varying driving scenarios.展开更多
基金fulfilled by Southeast University are supported partially by the National Natural Science Foundation of China under Grant 52402467 and Grant 52394263partially by the Natural Science Foundation of Jiangsu Province under Grants NO.BK20241324 and BK20233002+1 种基金partially by the"Southeast University Interdisciplinary Research Program for Young Scholars"The work fulfilled by Tianyi He are supported by Natural Science Foundation under award 1941524.
文摘This paper presents an Integrated physics-data-based(IPDB)modeling of lateral vehicle dynamics with moving-window data snapshots.The IPDB model encodes the fundamental physical principle of four-wheel vehicle motions and simultaneously carries out the adaptiveness of the data-driven approach.Specifically,the traditional physics-based lateral dynamics considering four-wheel interaction are first derived into an affine linear-parameter-varying model,in which the vehicle-related parameters and motion variables are separated.Then,by using the Kronecker product,the IPDB model,directly formulated by the data snapshots in the moving-window fashion,is obtained for system representation.As a result,the IPDB technique rendered model is physically interpretable.The impacts of moving window length on modeling performance are numerically studied.The IPDB model accuracy is validated with data from CarSim simulations and experiments with passenger vehicles under various scenarios.It is further demonstrated that the proposed IPDB model is more data-efficient than other data-driven methods since it only uses the data snapshot in the moving window to update the model recursively.This characteristic enables the IPDB method with online modeling capability to adapt to varying driving scenarios.
