Research On Modeling Method Of Rear-end Accident Risk Propagation Based On Risk Field Theory

Road traffic accidents have always been a major event threatening life safety and causing economic losses.Therefore,if the accident risk of upstream vehicles after the accident can be modeled and evaluated,it will be of great significance to the safety prevention of secondary accidents.However,most of the current assessment methods for vehicle accident risk do not clearly understand or directly ignore the risk propagation mechanism between vehicles,resulting in unscientific and reasonable accident risk modeling and inaccurate description of accident risk propagation characteristics,which reduces the efficiency of active safety prevention and control by drivers and road management departments after accidents.In order to solve this problem,the paper deeply analyzes the risk propagation mechanism between vehicles,proposes a rear-end collision accident risk assessment method combining the driving risk field and dynamic Bayesian network,establishes a rear-end collision accident risk propagation model,and studies the rear-end collision accident risk propagation characteristics.This research will provide technical support for vehicle anti-collision strategy formulation,risk management and control facility deployment and vehicle evacuation route planning under emergency events.The main work of the paper is as follows:（1）Analysis of the characteristics of rear-end collision accidents.The traffic accidents of Yongtaiwen Expressway are divided into 8 types according to the accident occurrence form.The statistical results further verify that rear-end collisions are the most frequent traffic accidents and the most serious traffic accidents causing personal injury and congestion delay.This is the premise of the research.（2）Analysis of risk propagation mechanism and proposed modeling method.The process of risk propagation among consecutive vehicles on a single-lane road is described.The analysis shows that rear-end collision risk of the rear vehicle should not only be related to the interaction risk of the front and rear vehicles,but also closely related to the risk of the front vehicle transmission.Based on this,a rear-end collision accident risk assessment method combining driving risk field and dynamic Bayesian network is proposed.（3）Construction of a rear-end accident risk propagation model.A driving risk field model suitable for the single-lane road scenario is established,and the vehicle interaction risk field force is converted into the risk probability,so as to construct a vehicle operation interaction risk model.Then a dynamic Bayesian network framework is built to describe the risk propagation between vehicles.Through the probabilistic reasoning process of the framework,the interaction risk of the front and rear vehicles is coupled with the risk of the front vehicle propagation,and a rear-end collision risk propagation model is proposed.（4）Experimental analysis of risk propagation model.The model parameters risk transfer probability,critical risk probability and critical risk field force are calibrated respectively,and the rationality of the established risk propagation model is proved by comparison.Finally,the propagation characteristics of rear-end risk with traffic state changes are analyzed from micro and macro perspectives.