Modelling,Simulation And Risk Evaluation For The Mixed Traffic Flow Of Autonomous And Human-Driven Vehicles

Autonomous vehicle technology is the product of the deep integration of the automotive industry with artificial intelligence,high-performance computing and other new information technology.It is the main direction of developing intelligent and connected automobiles and transportation.Road testing is necessary for the technology iteration and commercialization of autonomous vehicles.To this end,countries worldwide are pushing ahead with road tests of autonomous vehicles on a large scale.However,in the road test,the hybrid intelligent interaction between autonomous driving vehicles,whose technology maturity is still to be improved,and human-driven vehicles is easy to cause the oscillation of traffic flow,which leads to alternative "road rage" and other problems,forming a new type of mixed traffic flow with great risks and hidden dangers.To promote road testing safely and orderly,it is urgent to develop an effective risk assessment method.This study proposed a modelling,simulation and evaluation framework for mixed traffic flow of autonomous and human-driven vehicles in the road test phase.Based on a large number of realistic road test data,operation characteristics analysis and risky scenario identification of mixed traffic flow in the road test phase were carried out.We built a risk assessment system for mixed traffic flow.Several typical risk assessment combinations were selected to carry out risk assessment of mixed traffic flow based on realistic data and simulation.First,a simulation evaluation framework for mixed traffic flow was built.There are three main steps in the framework: operation characteristics analysis was carried out based on processed realistic data mixed traffic flow;after identifying the risky scenario,the mixed traffic flow under the risky scenario was simulated;build a traffic risk evaluation system for mixed traffic flow and carry out a traffic risk evaluation for mixed traffic flow to realize data-driven simulation modelling and risk evaluation for mixed traffic flow.Secondly,the processing requirements of multi-source data of mixed traffic flow were sorted out.According to different data processing requirements,a data processing framework of mixed traffic flow,including multi-source data fusion,abnormal processing of trajectory data and recognition of critical events in operation,was proposed,and specific data processing methods were given.We analyzed the operational characteristics,such as velocity,acceleration,and time headway,of humandriven vehicles,autonomous vehicles controlled by automation and autonomous vehicles controlled by human drivers in different scenarios and further analyzed the impact of scenario factors,such as road properties,traffic conditions on these operational characteristics.Thirdly,the LSTM method was adopted to identify the risk-avoiding disengagement of autonomous vehicles in the road test phase.Based on the Tobit model,the influence of scenario factors on the risk-avoiding disengagement rate was analyzed,and the risk scenario model was established.Based on the analysis results of the realistic data,the simulations of typical risky scenarios on urban roads and highways were conducted,respectively.Finally,a traffic risk evaluation system for mixed traffic flow was proposed,considering four dimensions: evaluation purpose,object,scenario and indicator.Based on practical application demands,typical risk evaluation combinations were given.The selection and construction methods of assessment indicators were proposed for different risk assessment combinations,including the selection methods of efficiency risk indicators,safety risk indicators and the modelling method of road test integration degree of autonomous vehicles considering multiple scenarios,multiple models and multiple indicators.The traffic risks of mixed traffic flow in different scenarios were analyzed based on simulation and realistic data.The research results provide a theoretical basis and technical support for the orderly,safe and controllable promotion of autonomous vehicle road testing and further promote the development of autonomous vehicle technology and the construction of intelligent transportation systems.