IJCAS

To obtain vehicle status information in real time, advanced algorithms have been proposed. However, the existing studies for vehicle systems only consider the accuracy of the estimation calculation and seldom consider the influence of the sensor data loss. In addition, the selection of the process noise covariance matrix (Q) and measurement noise covariance matrix (R) in the traditional Kalman algorithm is challenging and has an important influence on the performance of filter dynamic state estimation. To solve these problems, a new adaptive correction extended Kalman filter algorithm (ACEKF) and an adaptive timestamp extended Kalman filter algorithm (ATSEKF) are proposed, which can adaptively estimate the vehicle state when the sensor data are lost. The former uses the statistical characteristics of the packet loss variable, while the latter uses the timestamp technology to accurately obtain the value of the packet loss variable. Simultaneously, the two algorithms estimate R and Q through the innovation sequence and state equation, respectively. To make the estimated Q and R more accurate, fuzzy factors are added for adaptive regulation. Finally, based on a practical vehicle model, the effectiveness of the algorithms is proven by the joint simulation of CarSim and MATLAB, which demonstrates that the proposed algorithms are better than the classical extended Kalman filter. Because the timestamp technology is added to ATSEKF, the simulation results show that ATSEKF has better tracking effect than ACEKF.

Keywords: Kalman filter, packet dropping, sensor, timestamp, vehicle estimation.