Vehicle state and parameter estimation based on improved adaptive dual extended Kalman filter with variable sliding window

Abstract

Due to the complex driving conditions of vehicles and the inevitable measurement noise of sensors, it is challenging to estimate the state and road friction coefficient. Therefore, a vehicle state and parameter observer based on the improved adaptive dual extended Kalman filter for mechanical elastic wheels (MEW) is designed to improve estimation accuracy. Firstly, the nonlinear vehicle dynamics model matching MEW is established, and the time-varying longitudinal and lateral stiffness of the tire are calculated instantaneously. Then in order to deal with the time-varying measurement noise of the sensors, an adaptive adjustment strategy is adopted to update measurement noise covariance matrix in time according to the ratio of the traces between the theoretical measurement error covariance and the actual measurement error covariance. Afterwards a new adaptive adjustment strategy of sliding window is proposed, which is computed by time derivative of the measurement error, characterizing the correlation between the sliding window and the vehicle state change. Finally, through the co-simulation of CarSim and Simulink under different conditions, the conclusion shows that the observer not only ensures the better estimation accuracy of vehicle state and road condition, but also reduces the dependence of the observer on parameters.