State Estimation Of Vehicle Stability Control System Based On Adaptive Filtering

The electronic stability program(ESP) need more information about the state of the vehicle, such as tire longitudinal force, lateral velocity, horizontal angular velocity and sideslip angle. However, because of physical reasons or cost problems, some state information cannot be measured directly by the effective sensor. Therefore, the domestic and foreign scholars put forward a method, which based the state estimation model and state estimation method, to realize the estimation of vehicle state information, the method solves the problem of incomplete information.According to the need to consider the degree of freedom in the ESP control, choosing the seven degree of freedom vehicle dynamic model and using the CarSim simulation software to verify the model of seven degree of freedom vehicle dynamic model. Combined with the state parameter estimation theory, the dynamic equations of the model are reformulated as equation of state. According to the actual availability of sensors in the application of ESP, establish the measurement equation. Tire forces in the vehicle state equation and measurement equation are calculated through the tire model. From the control point of view, the random walk model is choose as tire model, which takes the tire force as unknown parameter to extend the state. Taking the vehicle state space model, whose state is extended, as the vehicle stability control system state estimation model.This paper introduces the advantages of classic Kalman filter(KF) theory. The classic Kalman filter is only suitable for linear systems, but the vehicle dynamics model is nonlinear,so,Kalman filter applied for nonlinear system should be adpopted. Therefore, on the basis of understanding the basic theory of the classic Kalman filter, this paper delves into the extended Kalman filter(EKF) and the Unscented Kalman filter(UKF) and the Adaptive Unscented filtering(AUKF), these filtering algorithm are based on KF and can be used for the nonlinear system state estimation. Among these filtering algorithm, EKF filtering algorithm use the Taylor formula to carried out the nonlinear function, so the nonlinear problem is transformed into a linear problem, but this method will produce the high order truncation error problem, and the amount of the jacobian matrix computation of the nonlinear function is relatively large.Due to the shortage of EKF filter, the UKF filter algorithm is raised. The filtering algorithm uses sampling method to approximate the nonlinear distribution to solve nonlinear problems. UKF is based on the UT transform and Kalman filter frame and it directly approximate posterior distribution by using deterministic strategies. UKF filtering process does not require derivative calculation of jacobian matrix, so the calculation time and the estimation accuracy of UKF is higher than EKF.EKF and UKF are based on the known statistical properties of the mathematical model and the accuracy of the noise. When the vehicle environment changes or motion state changes, the statistical characteristics of the noise will change greatly, which will cause EKF and UKF filtering accuracy and stability reduced greatly.Considering the car will encounter some interference in the process of moving, which has a certain impact on the vehicle sensor and leads to the statistical characteristics of the sensor is difficult to accurately obtain, that is to say, the characteristics of measurement noise is time-varying or unknown, Designing the AUKF algorithm. AUKF realizes the online estimation of the measurement noise covariance matrix through the information sequence, improving the adaptive ability of UKF. The simulation results show that when the measurement of noise is unknown or time varying circumstances, AUKF can estimate the state variables of automobile and has the best estimation accuracy.