Research On Soft-Sensing Algorithm Of Vehicle State For ESP

ESP(Electronic Stability Program) improves the active safety of vehicle significantly. It is a new vehicle active safety control system based on the ABS(Antilock Brake System)and TCS(Traction Control System), and has been equipped in many cars throughout the world. ESP is becoming an international hot spot of automobile active safety research. It not only integrates all the functions of traditional ABS and TCS, but can improve the stability of the vehicle under extreme driving situation.The entire ESP control process is based on the measurement and estimation of the vehicle motional state. However, because of different physical properties, the difficulty levels of measure these vehicle motional states are different. Even in the existing technology conditions and acceptable cost scope, some vehicle motional states can not be measured directly or indirectly. The measurement incompleteness of vehicle motional state brought the enormous difficulty to the development and popularization of ESP. Therefore it is the key technology in the research of ESP to measure and estimate the vehicle motional state precisely, as well as the premise and foundation to realize the ESP control task. At present, the yaw rate can be measured by a sensor directly. But other parameters, such as tire forces, vehicle longitudinal and lateral speed, coefficient of tire-road friction, can not be measured by sensors effectively and directly in conditions of existing test level and test cost. Therefore Bosch and other vehicle electronic control products suppliers usually establish some estimate formulas to solve the problem of the measurement incompleteness of vehicle motional state.In recent years, many scholars have developed a new approach of vehicle motional state estimation by using vehicle dynamics model and state estimation method. With this technical route and the project of global chassis control technology research project of State 863 Project, a soft-sensing algorithm of vehicle state for ESP is presented to realize the estimation of vehicle tire forces with vehicle multi-degree-of-freedom nonlinear dynamics model and nonlinear state estimation method, in which we can use software algorithm to replace the hardware sensors. The innovation of this paper is taking advantage of UKF algorithm which made it possible to calculate the nonlinear dynamics equation directly, without focusing on the Jacobian matrix. On the assumption that no available tire model, we can still estimate the tire forces as well as vehicle longitudinal and lateral speed synchronously. So it provides us with a new approach in the research of ESP and other advanced chassis electronic control system. The work of this paper is as follows:(1) After an analysis on the Kalman Filter theory, a soft-sensing algorithm is proposed based on 2-DOF vehicle dynamic model, and an off-line simulation can be made to validate the soft-sensing algorithm.(2) After an intensive study of nonlinear filter, such as EKF (Extended Kalman Filter) and UKF (Unscented Kalman Filter), a 7-DOF vehicle model is built including vehicle longitudinal and lateral motion, yaw and rotary motion of four wheels, laying a vehicle dynamics model foundation for the estimation of tire forces. Combining the above 7-DOF model and making full use of the sensors—including sensors of acceleration, steering wheel angle, yaw rate and wheel speed, we can establish a new method to estimate tire force, which is most difficult to be measured directly, and the minimum mean square error estimation of longitudinal and lateral speed, longitude and lateral tire forces can be obtained respectively.(3) Simulation test of the soft-sensing algorithm based on UKF. Selecting four representative driving situations and simulation results from state estimation and Carsim—an automobile dynamic simulation and analysis software—are compared. The comparison shows that the soft-sensing algorithm of vehicle state based on UKF obtains good results of estimating vehicle longitudinal and lateral speed as well as longitudinal and lateral tire forces. So the soft-sensing algorithm is proved effective.