Research On The Personalized Lane Keeping Assist Strategy With Driver Behavior Identification

Intelligent vehicle lane keeping assist(LKA)system is a typical lateral intelligent assistant driving system with lane departure warning,lane keeping assist control and other functions,which is conducive to reducing traffic accidents,improving driving safety and relieving drivers' workload.It is a relatively widely used system in advanced driver assistance system(ADAS).The current lane keeping assist system lacks comprehensive consideration of drivers with various driving style,so the designed lane keeping system will inevitably have problems such as poor adaptability and poor driving experience.Therefore,deeply studying on the evolution law of drivers' personalized driving behavior characteristics,establishing the representation and identification strategy of drivers' driving style,and making a comprehensive consideration of drivers' driving behavior is an important research direction of lane keeping assistant system in the future,and also the premise and foundation of realizing personalized advanced driving assistant system and personalized driverless vehicle.Supported on the National Key R&D Program of China(2016YFB0100904),National Natural Science Foundation of China(51775235),and Jilin Province Science and Technology Development Plan Projects(20170101138JC),this paper has carried out the research on intelligent vehicle personalized lane changing assistant system based on driving style.Based on the real vehicle driving data collection platform,the driver's driving style data is collected;based on fuzzy clustering algorithm and random forest algorithm,the driver's driving style representation and identification strategy is established;based on the vehicle motion model,the intelligent vehicle personalized lane departure warning strategy is designed.The cooperative control framework of lane maintenance is constructed,and the optimal curvature preview model is introduced to solve the optimal steering wheel angle of the vehicle.The codriving coefficient of lane keeping assistant system is set up by fuzzy control algorithm.Finally,the driver-in-the-loop platform is built and the simulation condition is designed for test and analysis.The main research contents of this paper include the following four aspects:(1)Characterization and identification of driving behavior of driversBased on the software and hardware of RT3002&RT-Range high-precision integrated navigation system,MicroAutoBox 1401,dSPACE ControlDesk,RTMaps and so on,a real vehicle driving data collection platform is built.Tested drivers are recruited from the society,and the driving style data of drivers are collected under typical working conditions.The fuzzy clustering algorithm is used to cluster drivers into three typical driving style: cautious,general and radical.Finally,the random forest algorithm is used to build the driver's driving style identification strategy.(2)Personalized lane departure warning strategy for intelligent vehiclesFirstly,this chapter trains a driving intention recognition model based on real vehicle driving data.On the basis of accurately identifying the driving intention of the driver,a lane departure warning model based on lane departure time is comprehensively considered.Based on statistical methods,taking into account differences in driving style of different drivers to improve the applicability of the system,a personalized lane departure warning threshold is designed to meet the individual needs of drivers with differentiated driving behavior for the lane departure warning system.(3)Personalized lane keeping assist strategy for intelligent vehiclesIn order to avoid man-machine conflict during the work of the lane keeping assist system as much as possible,a personalized lane keeping man-machine cooperative control framework is constructed,and an optimal curvature preview model is introduced to solve the optimal steering wheel angle which could bring the vehicle return to normal.For three different driving habits,the fuzzy controller is used to calculate the man-machine co-driving coefficients of the lane keeping assist system,so as to meet the needs of drivers with different driving style for the safety and comfort of lane keeping.(4)Test verification and analysis based on driver-in-the-loop platformFirstly,this chapter set up a driver-in-the-loop platform,construct virtual simulation scene and design test conditions,respectively test and verify the driver's driving behavior identification model,personalized lane departure warning strategy,and personalized lane keeping assistance strategy.The results prove that the intelligent vehicle personalized lane based on driving habits Maintain the accuracy,usability and achievability of assistive strategies.