Research On Driving State And Parameter Estimation Of Electric Vehicle Based On Information Fusion Technology

In recent years,the rapid development of the automobile industry has attracted more and more attention.Vehicle active safety control systems are widely used in commercial vehicles.The normal operation of these systems is based on the accurate acquisition of the dynamic state and parameters of the vehicle,such as vehicle speed,yaw rate,sideslip angle and mass,and tire cornering stiffness.As the key control objectives of vehicle stability control system,vehicle speed and sideslip angle can determine the running track of vehicle.In addition,as an important part of connecting the vehicle and the ground,the tire force formed by their interaction is the only driving force source of the vehicle,which can dominate the movement of the vehicle.Therefore,it is critical to know the vehicle driving state and tire force in advance for the vehicle active safety system.However,the above vehicle dynamics state cannot be directly measured on the mass production vehicle or the cost of measuring equipment is high.The traditional Kalman filter has good performance in estimating vehicle state parameters in Gaussian noise environment,but it shows poor accuracy and robustness in actual non-Gaussian noise.Vehicle model parameter perturbation,inaccurate tire force calculation and vehicle sensor signals are susceptible to non-Gaussian noise(especially heavy tailed noise),which pose great challenges to vehicle driving state parameter estimation.Therefore,the use of easy-to-measure vehicle sensor signals based on information fusion technology to estimate difficult-to-obtain vehicle driving state parameters has become a way to solve the above problems,which is also of great significance to the largescale production of vehicle control systems.This thesis focuses on the estimation of driving state parameters of electric vehicles.The main contents of this thesis are as follows:(1)The three-degree-of-freedom(3-DOF)and seven-degree-of-freedom(7-DOF)dynamic models of electric vehicles,the motor model and the nonlinear dynamic Dugoff tire model are established.Based on the Carsim-Simulink joint simulation platform,the accuracy of the established mathematical model is verified under sinusoidal conditions and double lane change conditions.(2)Based on the maximum correntropy criterion(MCC)and the adaptive extended Kalman filter(AEKF),the maximum correntropy adaptive extended Kalman filter(MCAEKF)algorithm is derived,and the computational complexity of the related algorithms is analyzed.A state parameter observer based on 3-DOF vehicle model is designed,and multi-condition simulation experiments are carried out through the joint simulation platform.The results show that MCC can effectively suppress non-Gaussian noise,and MCAEKF algorithm has stronger robustness and higher estimation accuracy than extended Kalman filter(EKF)and AEKF under non-Gaussian noise.(3)The vehicle mass is first identified by the least square method with forgetting factor(FRLS)and the lateral dynamics model.Secondly,based on the wheel rotation dynamics model and the single-track model,the adaptive sliding mode observer(ASMO)is used to accurately observe the longitudinal and lateral tire forces,and the mass and tire force information is called a pseudo-measurement.Finally,based on the obtained pseudo-measurement information,the maximum correntropy square-root cubature Kalman filter(MCSCKF)is used to estimate the driving state of vehicles including longitudinal speed,lateral speed and sideslip angle in non-Gaussian situation.Through two typical working conditions,it is proved that the proposed FRLS and ASMO can accurately estimate the vehicle mass and tire force.At the same time,under non-Gaussian conditions,MCSCKF shows high estimation accuracy and robustness for vehicle driving state estimation.Compared with the traditional estimation framework,the proposed robust hierarchical estimation framework can better deal with model parameter perturbation and non-Gaussian noise problems.(4)On the real vehicle test platform,a microcomputer steering wheel angle measuring instrument,a high-precision nine-axis attitude sensor,a centimeter-level high-precision GPS and a PC are configured.The test data is collected by the host computer,and the effectiveness of the proposed vehicle state estimation method is verified offline.