An Adaptive UKF for Vehicle State Estimation With Delayed Measurements and Packet Loss

The exact acquisition of vehicle states is a prerequisite to improve the safety of vehicles. However, existing methods for vehicle state estimation focus only on the improvement of estimation accuracy and rarely consider the effects of delayed measurements and packet loss of sensor data. To deal with this problem, an adaptive unscented Kalman filter with delayed measurements and packet loss is proposed for vehicle state estimation. Stochastic variables satisfying Bernoulli distribution are adopted to characterize the stochasticity of delayed measurements and packet loss. The mathematical expression of the adaptive unscented Kalman filter with delayed measurements and packet loss algorithm has been presented based on the theory of orthogonal projection. The gain matrix and covariance matrix of the estimation error are updated dynamically. Moreover, we have given a detailed proof of the closed-loop stability of the proposed algorithm. Simulation experiments and real vehicle tests indicate that the proposed algorithm has a higher estimation precision than existing ones that ignore the influence of delayed measurements and packet loss. Furthermore, the algorithm shows a strong robustness to different road conditions.