Estimation of Vehicle State Based on Limited Memory Random Weighted Unscented Kalman Filter

Critical state information of the vehicle is a prerequisite for effective implementation of active safety systems. Combining on-board sensor information and advanced filters to estimate vehicle states is much more economical than direct measurement methods. However, the influence of historical measurement data on the estimation accuracy is not fully considered in the existing estimation methods. Therefore, a novel limited memory random weighted unscented Kalman filter (LMRWUKF) is proposed to address this problem. First, a nonlinear three-degree-of-freedom vehicle dynamics model is established. Then, an iterative method to update the measurement noise covariance matrix using finite measurement data is embedded into the unscented Kalman filter to form the LMRWUKF. Finally, virtual tests for two different operating conditions show that the performance of LMRWUKF is better than the unscented Kalman.