FusionGrip: Tire model identification and stability control of 4WID vehicles under extreme conditions

Under extreme driving conditions, the handling performance and yaw stability of four-wheel independent drive (4WID) vehicles are significantly challenged by tire nonlinearities and force saturation. This paper proposes a hierarchical stability control framework, termed FusionGrip, based on a fusion tire model and optimization-based tire force vector control. To overcome the limitations of analytical tire models under combined slip, a high-precision fusion tire model is established by integrating the UniTire-Ctrl model with a BP residual neural network via Kalman filter, combining physical interpretability with nonlinear fitting capability. The model parameters are identified using real tire bench test data, and its generalization capability is validated under both training and untrained operating conditions. A BP neural network-based vertical load observer is developed to provide accurate real-time load estimation, enhancing tire force prediction fidelity. The fusion tire model and vertical load observer are applied to the hierarchical stability control framework, so as to achieve better performance. The hierarchical controller consists of three layers: driver intention interpretation, tire force allocation under friction circle constraints, and optimization-based tire force control that directly computes rear steering angles and wheel speeds, addressing the difficulties associated with coordinate transformation. Simulation results under double-lane change and aggressive racetrack scenarios demonstrate that the proposed strategy maintains vehicle stability with low sideslip angle and improved steering consistency, validating its effectiveness in enhancing both handling and stability limits under extreme conditions.