Study On Humanoid Steering Control Technologies Of Intelligent Vehicle Based On Experienced Driver Steering Manipulation Characteristics

In recent years,with the rapid integration of high-tech and advanced automotive technologies such as computers,the Internet,communications and navigation,automatic control,artificial intelligence,machine vision,precision sensors,and high-precision maps,smart cars(or driverless cars,self-driving cars)has become a research hotspot in the world of automotive engineering and a new driving force for the growth of the automotive industry.However,the intelligent vehicle’s automatic driving control technology is facing many difficulties about how to completely replace human drivers within the full range of operating conditions and to be accepted by customers.For intelligent vehicle,if the steering wheel angle is controlled only based on EPS(Electrical Power Steering)steering angle signals,and the output torque of the EPS motor is not controlled,steering wheel shake may occur.In the process of returning the steering wheel with a large steering angle,if the damping of the EPS motor(reverse torque)is not controlled,the steering wheel will be returned too fast and cause a large body yaw rate,which will degrade the quality of the steering control of the vehicle.The above conditions make the steering quality of the vehicle worsen,and even when the vehicle runs at a high speed,the control instability may occur,which endangers the driving safety.Therefore,how to make the steering control quality of the intelligent car in automatic driving mode as close as possible to the steering control of the human driver is an important issue in the research of intelligent vehicle steering control system.This paper analyzes the acquisition and characteristics analysis of the steering parameters from the experienced driver’s real vehicle road test,the non-linear fitting of the experienced driver’s driving trajectory,the construction of the intelligent vehicle driver model based on the MPC(Model Predictive Control),and the steering control strategies based on the human-simulated intelligent control theory and research on the development and experiment of intelligent vehicle humanoid steering control bench based on novel EPS device.The main tasks are as follows:The data collection and analysis of the road steering test for experienced drivers are carried out,5 experienced drivers are selected to perform real-vehicle road steering tests under different vehicle models,vehicle speeds and steering conditions.The acquisition parameters mainly include steering characteristic parameters and vehicle dynamics parameters;the interference signal is designed with Butterworth filter of different orders for filtering;through coordinate transformation processing,the latitude and longitude signals acquired by GPS/INS are converted into the two-dimensional plane of the vehicle trajectory in the geographic coordinate system;through analysis influence of various factors on the trajectory of the vehicle under different steering conditions and the relationship between the steering angle,the torque and the angular velocity,the experienced steering mechanism of the driver is obtained;the use of principal component analysis to determine the average steering angle and the average angular velocity as the most representative driver’s steering;according to different cluster analysis methods,the characteristics of the parameters of the driver were studied under different operating conditions,and the advantages and disadvantages of three different clustering methods(FCM,GK and GG)on the classification of driver steering characteristics were compared and analyzed.GG algorithm is determined as the optimal clustering algorithm,and The clustering center of the driver’s steering characteristics is also determined under different steering conditions;threshold value of the experienced driver steering parameter is obtained using the above clustering center,and it lays the foundation for the design of the human-simulated intelligent control law.The nonlinear fitting method of the experienced driver’s trajectory is studied.Segmented polynomial expressions are applied to four typical steering conditions including right-steering,U-turn,lane-keeping,and lane-changing;for traditional neural network learning algorithms(such as BPNN),artificially setting a large number of network training parameters is required.The disadvantages of local optimal solution are easy to generate.Using extreme learning machine(ELM)has the advantage of fast learningand good generalization performance in nonlinear fitting.An ELM-based method for non-linear fitting of experienced driver driving trajectory is proposed.Using Kalman filter(KF)algorithm to update the output weight of ELM in real time.It solves the problem of low learning accuracy of ELM in the case of multicollinearity,and filter the ELM output weight matrix and update it in the update phase to achieve optimization of the ELM algorithm;using KFELM,ELM,and BPNN to perform non-linear fitting tests on the experienced driver’s driving trajectory under different operating conditions.The experimental results show that KFELM’s training accuracy and test accuracy are significantly better than those of ELM and BPNN.At the same time,KFELM’s learning speed is slightly better than that of ELM,and it is obviously better than BPNN.The MPC algorithm is used to predict the future state of the system and perform rolling optimization at each sampling time,and a driver model based on MPC is constructed;according to the requirements of the vehicle kinematics model to quickly and steadily follow experienced driver driving trajectory,the objective function of the driver model and the corresponding constraints are determined.Through the simulation analysis,this paper compares the MPC-based driver model with the conventional single-point preview driver model and the path tracking method based on β-spline curve.The results show that under the four typical steering conditions,the MPC-based driver model can accurately track the experienced driver’s reference trajectory,and tracking effect is better than two traditional methods.The design steps of the human-simulated intelligent controller based on the kinesthetic schema are designed.The concept of the new EPS steering system based on the dual torque/rotation angle sensor is proposed.The dynamic model of the steering system is established,and the relationship between steering resistance moment and speed and steering angle is analyzed.human-simulated intelligent control law based on kinesthetic diagrams was established.Segmented control of EPS steering angle and torque is presented;humanoid steering control system of intelligent vehicle based onMTLAB/Simulink is constructed,and the joint simulation test platform based on Carsim/Simulink is bulit.Simulation results show that human-simulated intelligent control algorithm is superior to traditional PID control in the tracking effect of steering characteristic parameters and passenger comfort.A steering resistance torque simulator based on magnetic powder brake and an intelligent vehicle steering system test bench were set up.The dSPACE rapid control prototype was used to perform human-simulated intelligent control.The actual realization of the law,humanoid steering control system test bench is established;test analysis of system control performance is conducted.The test results show that the steering angle obtained by the test bench can roughly track the real vehicle test value of the experienced driver,especially under the lane-keeping operating condition,the maximum steering angle deviation is less than 2deg,and a ideal test effect is obtained;about the steering torque,In terms of moments,The test results show that the steering angle obtained by the test bench can roughly track the real vehicle test value of the experienced driver,especially under the lane-keeping operating condition,the maximum torque deviation is less than1 Nm.The human-simulated intelligent steering control system proposed in this paper has completed the task of human-like driving well,and the stability and effectiveness of the designed controller is verified,as well as the control performance of the new EPS steering system.The research shows that the steering angle and torque of the steering wheel can preferably track the steering angle and torque curve of the actual driver steering test of the excellent driver.The maximum steering angle deviation in the lane-keeping operating condition is less than 2 deg and the maximum torque deviation is less than 1 Nm.The effectiveness of the humanoid steering control technology studied in this paper is verified.