Abstract: The driving mode of hybrid electric car can be two-wheel or four-wheel drive. The speed estimation designed for a specific driving mode can hardly achieve high estimated accuracy in all modes. Focus on the changeable driving mode, a kind of distributed Kalman filter with two levels is proposed, by means of utilize vehicle sensors signals, the driving torque of rear in-wheel motors and the torque of front driving wheels at the given gear state. Considering the strongly nonlinear of the model, UKF is adopted to design the filter, and to improve the filter’s robustness on modeling error and signal noise, measurement noise mean and covariance is self-adaptive. Firstly, based on UniTire Model and recursive equation of slip rate, the sub-filter is developed; secondly, based on the vehicle Kinematics model, the master-filter can be developed; at last, in consideration of different driving modes, a more accurate velocity estimation is obtained by fusing information from each sub-filters. With the help of Carsim and Simulink, the simulation platform about four wheel drive hybrid electric car is developed, and the proposed speed estimation algorithm is tested on this platform under the pure electric drive mode and the hybrid drive mode. The results show that the proposed algorithm has not only high precision, but also strong robustness on modeling error, measurement noise and the changeable driving mode.

Key words: distributed filter structure, four-wheel drive hybrid electric car, self-adaptive unscented Kalman filtering algorithm, vehicle speed estimation