Abstract: To accurately obtain the longitudinal velocity of the vehicle, a longitudinal velocity estimation method applicable to four-wheel drive vehicles is proposed. Firstly, a finite state machine is utilized to identify the vehicle state at the current moment and the vehicle state in the time-domain window, which effectively switches between the adaptive Kalman filtering method and the integration method. For the four-wheel non-total skidding state, an adaptive Kalman filter method that updates the measurement noise in real time is designed. This method introduces the measurement value and estimation error in the time-domain window to improve the estimation accuracy. For the four-wheel total skidding state, the last longitudinal velocity estimate from adaptive Kalman filtering is used as the initial value, and the longitudinal velocity is calculated by integrating the longitudinal acceleration of the vehicle. The effectiveness of the algorithm is verified by Carsim and Simulink joint simulation experiments and real vehicle data experiments. The experimental results show that the estimation accuracy of the proposed estimation method is improved by at least 65% and 75% on low-adhesion road surfaces such as snow and ice, respectively, compared with the integral method and the method of estimating longitudinal velocity using wheel speeds.

Key words: longitudinalvelocity estimation, adaptive kalman filter, four wheel drive vehicle, finite state machine, time domain