Vehicle State Observation and Road Adhesion Coefficient Estimation based on UKF

A cascaded Unscented Kalman Filter (UKF) framework is proposed for the joint estimation of vehicle states and road adhesion coefficients. First, a three-degree-of-freedom (3-DOF) vehicle dynamics model is derived and a UKF-based observer is designed to fuse wheel-speed, acceleration, and yaw-rate measurements for real-time estimation of longitudinal and lateral velocities and yaw rate. Next, the estimated states are fed into a nonlinear Dugoff tire model and the resulting tire forces are normalized to eliminate the dynamic influence of vertical load. Based on the mapping between normalized tire forces and vehicle motion states, an observation equation is formulated with the adhesion coefficients of all four wheels as state variables, and the UKF is applied for their dynamic identification. Theoretical analysis demonstrates that the cascaded structure prevents error accumulation typical of sequential estimation and suppresses sensor noise in strongly nonlinear systems via sigma-point propagation. Finally, co-simulation in CarSim and MATLAB/Simulink confirms the method's accuracy and robustness, providing a solid theoretical foundation for vehicle active safety control and multi-source perception in autonomous driving.