Analysis Of Driving Behavior And Assessment Of Early-Warning Strategies In Rear-End Collision Risk Scenarios On Urban Roads

Car-following is a common driving state,and rear-end collision is one of the common accident types,which accounts for a large proportion of road traffic accidents,seriously threatening people’s life and property safety.From the perspective of accident mechanism,driver’s risk behavior is the main and direct cause of rear-end collision.Vision is the main way for drivers to obtain driving information,the visual perception and driver’s mental activities directly affect driving behavior.In this study,the cases of rear-end collisions in China in-depth accident study database were selected.The light gradient boosting machine and SHAP algorithm were combined to construct the prediction and characteristic analysis model of driver accident injury severity.The influence of accident characteristic parameters on driver injury severity was analyzed.Based on the analysis and investigation results of rear-end accident causes,the representative rear-end collision risk scenario was constructed.The indicators of driver’s reaction behavior under the influence of adverse visual interference factors were studied through driving simulation and questionnaire survey.Then,the relationship between driver’s reaction behavior indicators and collision risk under high-risk scenarios was analyzed.Lastly,the active safety system efficacy of warning strategies was evaluated.This study is of great significance for improving the driving safety of urban road vehicles,reducing the occurrence of rear-end collision accidents,and improving the level of road traffic safety.First of all,the secondary task experiment of in-vehicle information system based on visual distraction was constructed under the car following state,and the indicators of the driver’s reaction behavior and eye movement characteristics were collected.Statistical methods such as descriptive statistics and multivariate analysis of variance were used to explore the driver’s differential performance under the influence of the secondary task of in-vehicle information system.The factor analysis method is adopted to build a quantitative model of driver’s mental workload based on driving behavior and eye movement indicator.And the effectiveness of the model was verified by the NASA-TLX subjective task load scale.At the same time,in order to facilitate the identification of drivers’ mental workload,a driver’s mental workload identification model was constructed based on the Bayes discrimination method to realize the accurate identification of driver’s mental workload of visual distraction secondary tasks under the car following state.Secondly,to explore the influence of different visual visibility(foggy environment)and road alignment interaction effects on driving behavior,good visibility(fog free environment)was taken as a comparison,and the linear mixed model was used to fully consider the individual behavioral differences of drivers and explore the influence of visual visibility on driving behavior.Then,the difference of driver’s collision avoidance reaction behavior under the influence of visual visibility was explored.At the same time,considering the influence of visual visibility on the driver’s operating behavior characteristics of the curve,the probability model of vehicles driven out of the curve was constructed based on the Logistic regression method,to predict the probability of vehicles driven out of the curve.Thirdly,based on different TTC values triggering the leading vehicle braking,the collision avoidance behaviors of drivers were analyzed under rear-end collision conflict conditions,which were affected by the degree of collision risk and the running state of the leading vehicle.From the perspective of driver’s visual urgency,based on the non-parametric test method,the correlation between driver’s visual looming and braking behavior was analyzed.The risk prediction model of driver rear-end collision was constructed to predict the collision probability of different driver types under the influence of the rear-end collision risk.Different warning strategies have different effects on drivers’ reaction behavior,an experiment for the efficacy of early warning strategies was constructed to collect driver’s driving behavior characterization parameters.The multivariate analysis of variance was used to explore the differential impact of early warning strategies on driving behavior.Based on the factor analysis method,a quantitative evaluation model was constructed to compare and evaluate the safety efficacy of early-warning strategies.The statistical model of driver’s reaction behavior under the effect of early warning strategy was constructed based on the simulated driving test.The counterfactual simulation method was used to combine the driver’s reaction behavior model and forward collision warning algorithm into the vehicle’s pre-crash motion state in the rear-end collision reconstruction case.The efficacy performance of drivers in avoiding collision conflicts and collision injuries under the influence of early warning strategy was explored.And the driver’s behavior under the rear-end collision risk scenario is systematically studied,based on the comparative evaluation of the safety efficacy of the warning strategy.Based on the above research,the driver’s reaction behavior characteristics in rear-end collision risk scenarios on urban roads are deeply analyzed.The study results enrich the theories related to rear-end collision behavior,provide a theoretical basis for exploring the formation mechanism of rear-end collision accidents and training of scientific driving skills,and provide a reference for the development and efficacy evaluation of active safety system warning strategy,which is conducive to improving driving safety.Thus,effectively preventing and reducing the occurrence of rear-end collision accidents.