In order to get the accurate and real-time vehicle state variables in running, a new kind of model adaptive update particle filter method is proposed. The non-Gaussian and non-linear tire noise vehicle dynamics model is established. High frequency sub-band is used to estimate real-time measurement noise variance of sensors based on the wavelet transform. The real fitting degree of observation likelihood function is improved and the degradation phenomenon of particle weight is improved to a certain extent by the combination of the adaptive auto regression model of the whole vehicle system state. Virtual experiment based on ADAMS/Car and real vehicle experiment verify the validity of the proposed method. Experiment results show that the estimation precision and anti-noise performance of the proposed method are superior to those of the commonly used method, and can satisfy the requirements of vehicle state estimation.