Tire Critical States Estimation Of Intelligent Vehicle

Whether it is vehicle stability control for vehicle safety or motion control for intelligent driving,real-time and accurate acquisition of vehicle system state variables is crucial.Tire mechanical properties are the basis of vehicle dynamics,and the estimation of vehicle state largely depends on the state of four wheels.For high-level automatic driving technology that lacks human drivers,the state estimation module ensures the stable operation of the active safety system downwards,and inputs decision planning and motion control modules upwards as ego of vehicle constraints and control objectives.To this end,this paper is committed to designing a reliable key states observation system for intelligent vehicle tires,which can accurately estimate the vehicle state including the tire mechanical states,kinematic states and road friction coefficient,to improve the comprehensive perception and real-time monitoring capabilities of the intelligent vehicle control system on the state of the vehicle,as also improve the vehicle control performance.The specific research contents are as follows:Firstly,based on the intelligent electric vehicle platform,a joint observation system for the longitudinal force,lateral force and vertical force of each wheel tire is designed.With the help of wheel kinematics,PID observer and Kalman observer are respectively designed to directly observe the tire longitudinal force,and the observation performances of the two are compared.Based on the three-degree-of-freedom dual-track model,the tire load observer is designed according to the load transfer after decoupling the sprung mass and unsprung mass.Then,taking the above observations of tire longitudinal and vertical force as input,an unscented Kalman filter for tire lateral force estimation is designed with vehicle dynamics.Through simulation experiments,it is verified that the proposed tire force estimation system is completely independent of the tire model and road friction coefficient,and can cover the tire force estimation problem of longitudinal slip and cornering conditions under different road conditions,the estimation results have high accuracy,which can provide reliable state information for subsequent observers.Then,for the intelligent vehicle equipped with GNSS system,a combined speed observer based on vehicle dynamic response and GNSS satellite system is designed to observe the kinematic state of each tire.Considering the designed requirments of the observer,the traditional high-precision multi-parameter UniTire model is simplified to a functional model that can express the tire force under pure and combined working conditions only relying on nine parameters.Based on the simplified UniTire tire model,a tire velocity Kalman filter in dynamic category is designed to estimate the velocity,slip rate and slip angle kinematic state of each tire.Under the influence of sensor characteristics and environment,the acceleration signal used by the observer will carry noise and offset bias,considering the Kalman filter’s ability to deal with Gaussian noise,an acceleration offset steady-state compensator is designed.With the continuous improvement of GNSS technology,the accuracy and reliability of GNSS system observations have made great progress.The above speed estimation Kalman filter is fused with GNSS observation based on the GNSS signal status flag to form a high-precision intelligent vehicle speed observation system suitable for all working conditions.The proposed observer can automatically degenerate into a pure dynamic velocity estimator when the GNSS data is not updated,ensuring the normal operation of the entire observation system.Finally,for the problem of road adhesion identification,a road surface friction classification identification algorithm based on forward camera and a forward uniform road friction level observer based on online vehicle/tire model are designed.Considering that the traditional method based on dynamic response is difficult to accurately estimate the dynamic value of road friction coefficient under small excitation conditions.On the one hand,the visual sensor is introduced,taking the camera signal as input,the convolutional neural network VGG16 is improved,and a road classification recognition algorithm for feature extraction and recognition of three-channel road image information is proposed.On the other hand,the realtime numerical monitoring of road parameters is abandoned,and a tire working area is used as an entry point to design a forward and uniform road friction level observer.When the tires are in the linear region,the dynamic upper and lower limits of road friction are estimated,and when the tires are in the nonlinear region,the road friction coefficient level is identified,to provide road surface perception information for active safety systems and intelligent driver assistance systems to optimize the control strategies of each system and improve the maneuverability of chassis actuators.