Research On Vehicle States Estimation Based On Adaptive Filter

With the increase in the production of vehicles, the safety issues of vehicles on road become thefocus of our attention, thus making active safety technology become a research hotspot. Automotiveelectronic stability control system (such as the Bosch’s ESP) is one of the important active safetytechnologies; It takes active intervention to the vehicle which is in an ultimate instability state toensure its safety driving through a certain control strategies. Such a control systems require accuratestates of the current vehicle to make a judgment of whether the vehicle is instability, and then adetermination will be made that whether the active intervention is applied.The vehicle states are the vehicle’s speed, yaw angular velocity, sideslip angle and otherparameters; As important information, some of the parameters may be more convenient to access, butsome parameters (such as the sideslip angle) requires expensive sensors to measure, and this hascaused great difficulties for the development and production of electronic stability control system;Currently, these parameters are obtained in estimation algorithm, so to study how to access theseparameters in a low-cost and accurate way has a great significance for the production of stabilitycontrol system.This paper summarizes the latest research states at home and abroad, then innovation andimprovement of the existing theories is carried out, the main work of this paper is as follows:1The Kalman filter algorithm is first introduced briefly, and then the emphasis is put on theclassification of the adaptive Kalman filter. At the same time, two main lines in this paper areelicited, that is the adaptive of kinetic model parameter and algorithm parameter. In order toachieve the adaptive of dynamic model parameters, RLS method is introduced, and then a serialRLS method for dual parameter identification is put forward.2According to the first main line, EKF and RLS is combined, the states estimated using EKF arepassed to the RLS algorithm for the identification of vehicle mass. The parallel estimation of thetwo algorithms can achieve online adjustment of vehicle mass when it changes, thus achieving theadaptive of vehicle mass under different load cases. The better result of the parallel algorithm isverified through virtual test.3According to the second main line, the measurement noise covariance matrix Fuzzy controller(FKF) is designed in Matlab fuzzy toolbox bsaed on fuzzy logic, and then statistical properties oftime-varying noise are estimated online. Then, an algorithm-S-AKF algorithm is introduced. FKF and the S-AKF are combined together, the adaptive of observation noise is achieved, at thesame time, the algorithm improves its robustness for the process noise. The validity of thealgorithm is verified through virtual test.4Vehicle stability control system is designed based on the yaw rate and the sideslip angle controlstrategy in PID, and Estimated–Control closed-loop and feedback stability control system isdesigned. Finally, the active intervention of the car in certain conditions is achieved throughco-simulation of Carsim and Simulink.