Identification And Control For Driving Risk Of Vehicle Using Naturalistic Driving Data

Vehicles,as the society’s most important means of travel and transportation,have a constant effect on our life and work.Traffic accidents remain numerous as more and more vehicles are put into circulation,and therefore improvement in vehicular safety has become an urgent task.Driving behavior is a factor closely related to vehicular safety and road traffic safety.Moreover,the complicated driving situation in China creates an opportunity for a profound study in safe driving,and deep learning based artificial intelligence has become a focus in safe driving assistance.Driving data understandably have an important role to play in pushing for research in this field.On the other hand,the building technologies for naturalistic driving platform are maturing,and it becomes a likely possibility to conduct an in-depth investigation into driving risks during vehicle travelling through acquisition and analysis of naturalistic driving data in real traffic situations.Naturalistic driving experiments have started in both China and other countries.The types and amount of naturalistic driving data are far above a level that renders it possible to make technical analyses,however with the some challenges to be overcome,such as naturalistic driving data encoding,driving risk clustering,driving risk identification,and active driving safety technology.The study presented in this paper,while having its source of research based on naturalistic driving data that have already been accumulated,includes a range of efforts from encoding of naturalistic driving data,mining of naturalistic driving data,identification of driving risks,and active safety control.The focus of the work is:(1)By following the idea of time series symbolization,naturalistic driving data are encoded in spatial and temporal semantics according to their features and based on symbolic aggregate approximation(SAX);techniques that try to analyze driving safety through encoding of naturalistic driving data are investigated from three perspectives: video clip capturing,time domain feature of driving data,and Haddon matrix description.(2)Macroscopically,the driving risks from multiple sample vehicles in a long time interval are classified and clustered using data mining technology to accurately capture the vehicles with higher driving risks,and a theoretical approach is proposed that effectively joins the relationship among vehicle driving data,driving behavior,and driving safety.(3)Microscopically,driving risks of individual vehicles during a short time interval are identified using data driven technology,and driver factors,vehicular movement status,and road condition are accounted for while considering the influence on vehicular traffic safety;belief rule base(BRB)is used to build a vehicle driving risk identification model that considers above multi-sourced and heterogeneous driving parameters and the model is implemented in a driving simulator situation.(4)In the future intelligent vehicle networking situation,active anti-collision control strategy for vehicles in a complicated traffic condition is investigated using model predictive control(MPC)based on evaluation of the risks posed by the vehicle on nearby traffic environment;the proposed active anti-collision control strategy is tested for its functionality in a specified complicated traffic simulation condition.This paper makes full use of the available accumulated naturalistic driving data and studies the technique that encodes and analyzes naturalistic driving data,and relevant work is carried out with the emphasis placed on clustering and identification of vehicle driving risks in the hope of finding a way to naturalistic driving data driven vehicle driving risk study.Finally,active anti-collision control strategy based on driving risk perception is explored in the future intelligent vehicle networking situation.The findings are of both great theoretical significance and broad application potentials to the supply of highly intelligent driving service,the improvement in driving safety,and the achievement of highly automatic driving in the future intelligent vehicle networking environment.