Data-driven physics-informed adaptive weighted fusion framework for modeling tire longitudinal force-slip characteristics

Traditional tire mechanism models often fail to accurately capture the complex dynamic characteristics of the tire-road interface, whereas purely data-driven models suffer from insufficient physical consistency. To address these issues, this paper proposes a data-driven physics-informed adaptive weighted fusion (DP-AWF) method for modeling tire longitudinal force-slip characteristics. The proposed method first employs a wavelet-function-enhanced Kolmogorov-Arnold Network (KAN) to learn the nonlinear mapping among slip ratio, vertical load, and longitudinal force. A LuGre dynamic friction model is then introduced to describe the evolution of friction during tire-road contact. Finally, local prediction errors of the two models under different operating conditions are statistically evaluated on an independent calibration dataset, continuous error functions are then constructed, and the fusion weights are adaptively adjusted according to the current operating condition, thereby achieving complementarity between data-fitting capability and physical constraints. Validation results based on emergency-braking simulation data for high and low adhesion road surfaces constructed in CarSim demonstrate that the proposed model can more accurately describe the tire nonlinear longitudinal force-slip characteristics. Compared with the Magic Formula and multilayer perceptron models, the proposed model reduces the root mean square error by approximately 69.2%–70.3% on average. Compared with a fixed-weight fusion method, it reduces the root mean square error by approximately 12.4%. The results indicate that the DP-AWF model combines the nonlinear fitting capability of the data-driven model with the dynamic trend constraint of the physics-based model, improving prediction accuracy while enhancing model stability and physical plausibility. This study provides a reference for tire-force estimation in vehicle braking control and active safety systems.