Research On The Identification Method Of Driver Steering Behavior Characteristics

With the popularity of X-by-wire technology applied on automobile,many traditional mechanical/hydraulic systems of automobile are being replaced by electronic sensors and actuators,which greatly promotes the process of automobile electronization,intelligentization and electrification.Meanwhile,the control algorithm design of automobile can be flexible,and the control parameters can be adjusted arbitrarily.As a result,there is much more design freedom on the design of dynamics control for X-by-wire automobile than for traditional automobile.On the premise of fully understanding driver Characteristics,the current situation of “driver adapts to car” can be changed to “car adapts to driver” to realize the ideal vehicle response by the personalized design of automobile dynamics control.The driver is the weakest and the most uncertain link in the driver-car-environment closed-loop system.By analyzing driver behavior characteristics,the deep reasons of road traffic accidents can be explored.And then,the corresponding measures can be taken to improve road safety performance fundamentally.It is necessary to consider the mutual adaptation between driver assistance systems and drivers.The research of driver behavior characteristics is good for improving the design of driver assistance systems and improving drivers' acceptance toward driver assistance systems.Driverless car is the trend in the development of automobile technology,and the essence of driverless control strategy is to imitate outstanding human drivers driving the car safely,efficiently and comfortably.Therefore,the reasonable design of driverless control strategy should be based on detailed research on driver behavior Characteristics.From the above all,it is necessary to research on the driver Characteristics from the aspect of personalized design of dynamic control for X-by-wire automobile,improving road safety performance from the perspective of drivers,humanized design of driver assistance systems and human-imitating intelligent control of driverless car.This paper combines itself with the National Natural Science Fund Project “Research on the reconfigurable integrated control strategy of distributed drive-by-wire electric vehicles”(Number: 51505178)and the National Natural Science Fund Project “Research on the control mechanism and evaluation method of the novel steering-by-wire system based on the driver's characteristics”(Number: 51575223).This paper focuses on the research of driver steering behavior characteristics.Feature parameters are obtained from the virtual turning experiment designed on the driving simulator experimental platform.The identification model of driver steering behavior characteristics is established based on the experiment data with the aid of pattern recognition theory and data mining technology.The main work is as follows:(1)Research the effect of external environment condition on driver steering behavior characteristicsThe purpose of this study is to identify the driver's inherent steering behavior characteristics through the driver's steering operation behavior and vehicle's movement state variables under the particular steering condition.While the shown driver's steering behavior Characteristics are affected by external environment parameters(vehicle factors,road conditions,traffic conditions and weather conditions,etc.)and the driver's inherent steering behavior Characteristics together at the same time.Therefore,it is necessary to identify and eliminate the effect of external environmental factors on driver's shown steering operation behavior characteristics before developing the identification model of driver steering behavior characteristics.Left/right turning orthogonal experiments are designed and implemented to examine the effect of the road conditions(turning radius,the turning angle of road,road width and tire-road friction coefficient)on the shown driver steering behavior characteristics.The left turning orthogonal experiment and right turning orthogonal experiment are designed and implemented respectively to explore the difference between driver's left and right steering behavior characteristics.(2)Classify driver steering behavior characteristic reasonablyTurning experiment on the driving simulator experiment platform is designed after fixing the road conditions(turning radius,the turning angle of roads,road width and tire-road friction coefficient).A number of drivers are selected to take the turning experiment.The driver's steering operation behavior and vehicle's movement state variables under the particular steering condition are recorded.Reference to previous research results,the optimal classification number of driver steering behavior characteristics is determined combining with related data mining technology.The driver steering behavior characteristics are clustered based on K-means algorithm,the Gaussian Mixture Model(GMM)and the Expectation Maximization(EM)algorithm.(3)Establish the identification model of driver steering behavior characteristicsBP Neural Network(BP_ANN)and Support Vector Machine(SVM)are two typical and effective methods to solve the pattern recognition issues.Identification model of driver steering behavior characteristics is established based on the right turning experiment data with the aid of the above two methods respectively.By comparing their test accuracy on the test samples,it is concluded that the test accuracy of BP Neural Network(BP_ANN)identification model of driver steering behavior characteristics is higher than that of Support Vector Machine(SVM).Finally,the prediction ability of BP Neural Network(BP_ANN)identification model of driver steering behavior characteristics is verified.