Vehicle state estimation by unscented particle filter in distributed electric vehicle

Abstract

Focusing on the existing problems in vehicle state estimation, such as deficient integrality system and low accuracy, unscented particle filter (UPF) is proposed, which utilizes the characteristic that distribute electric vehicle (DEV) has multi-information sources. Based on the nonlinear vehicle dynamic model, the novel system estimated longitudinal velocity, sideslip angle, yaw rate and tire lateral force simultaneously. In order to improve accuracy of tire lateral force, nonlinear dynamic tire model is utilized. UPF is developed according to the proposed vehicle and tire model. To improve the accuracy of UPF, measurement noise covariance is also self-adaptive regulated. Simulations and experiments show that the proposed method can improve robustness and accuracy of vehicle state estimation. © 2013 Journal of Mechanical Engineering.