Residual Learning towards High-fidelity Vehicle Dynamics Modeling with Transformer
Jinyu Miao, Rujun Yan, Bowei Zhang, Tuopu Wen, Kun Jiang, Mengmeng Yang, Jin Huang, Zhihua Zhong, Diange Yang
TL;DR
This work tackles the challenge of achieving high-fidelity vehicle dynamics for autonomous driving by combining physics-based baselines with a Transformer-based residual-correction network (DyTR). DyTR learns state residuals conditioned on historical states, control actions, and vehicle configuration, using a temporal Transformer to fuse dynamics features and a residual-query mechanism to iteratively refine predictions. Evaluations on co-simulated data for distributed electric-drive vehicles show that DyTR delivers substantial accuracy gains over both physics-based and generic DNN approaches, achieving up to 92.3% mean-error reduction on one dataset and 59.9% on another. The approach offers a practical path toward reliable long-horizon state estimation, enabling more accurate planning and control in high-fidelity vehicle dynamics scenarios.
Abstract
The vehicle dynamics model serves as a vital component of autonomous driving systems, as it describes the temporal changes in vehicle state. In a long period, researchers have made significant endeavors to accurately model vehicle dynamics. Traditional physics-based methods employ mathematical formulae to model vehicle dynamics, but they are unable to adequately describe complex vehicle systems due to the simplifications they entail. Recent advancements in deep learning-based methods have addressed this limitation by directly regressing vehicle dynamics. However, the performance and generalization capabilities still require further enhancement. In this letter, we address these problems by proposing a vehicle dynamics correction system that leverages deep neural networks to correct the state residuals of a physical model instead of directly estimating the states. This system greatly reduces the difficulty of network learning and thus improves the estimation accuracy of vehicle dynamics. Furthermore, we have developed a novel Transformer-based dynamics residual correction network, DyTR. This network implicitly represents state residuals as high-dimensional queries, and iteratively updates the estimated residuals by interacting with dynamics state features. The experiments in simulations demonstrate the proposed system works much better than physics model, and our proposed DyTR model achieves the best performances on dynamics state residual correction task, reducing the state prediction errors of a simple 3 DoF vehicle model by an average of 92.3% and 59.9% in two dataset, respectively.