Research On Pedestrian-Vehicle Collision Avoidance Behavior And Warning Effect In Visual Blind Area Based On Driving Simulation

Pedestrian-vehicle collisions are one of the accidents that often occurs in road,which often causes a large amount of property losses and serious casualties.Among all the reasons for causing pedestrian-vehicle collisions,collisions caused by visual blind spots are the most difficult to avoid,commonly known as "ghost probes".The visual blind spots are formed due to side-parking,buildings,bus stops,flyover,green belts and other road environments which block the driver’s sight.When pedestrians cross the road in front suddenly,the driver cannot obtain the information and take measures in time to avoid pedestrian-vehicle collisions,so it is necessary to reduce the incidence of such collisions.Based on the theoretical research of driver’s driving behavior in collision avoidance,side-parking section was selected as the experimental base scene,driving simulator was used to carry out the experiment.After collecting the driver’s driving behavior data and extracting key variables,descriptive statistics,mixed-effects model,binary logistic model and other statistical methods were used to analyze driving behavior and warning efficacy.Then,a risk evaluation model for non-collision events and a pedestrian injury degree prediction model for collision events were established respectively.Lastly,combined with the questionnaire survey,the difference between the drivers’ real driving behavior and his/her subjective consciousness was compared and suggestions for improvement of warning technology were put forward.The study found that the speed limit and warning release time have significant impacts on driving behavior and collision rate.In the collision avoidance process,as the speed limit level increases,drivers tend to choose to brake with greater deceleration and combine steering operations to avoid collisions.The rate of deceleration is smaller and the rate of deceleration growth rate is greater without warning released.On the other hand,as the time of warning early released,the brake reaction time,the maximum deceleration rate and the rate of deceleration growth rate all showed a downward trend.In terms of collision avoidance results,the speed limit,pedestrian speed,and warning release time have significant impacts on the collision rate.It was found that the collision rate was the lowest when the warning was released 4.0 seconds in advance.In addition,with the advancement of the warning early released,the minimum distance between pedestrians and vehicles showed an increasing trend.In order to better understand the impacts of various factors and key variables on pedestrian-vehicle collisions,two models were established.For non-collision events,TTC was selected as the evaluation index,and the collision risk was divided into three levels: high,medium and low risk.For collision events,the collision speed was selected as the predictive index,and the degree of pedestrian injury was divided into three levels:minor injury,serious injury and death.Logistic regression was used to establish the pedestrian-vehicle conflict collision risk evaluation model and the pedestrian injury degree prediction model.Both of the two models have high prediction accuracy.Finally,according to the questionnaire filled out by the drivers after the experiment,it is found that drivers have a high degree of satisfaction with the authenticity of the driving simulator.The factors influencing the authenticity evaluation of driving simulator were analyzed by using linear regression method and there is a significant linear relationship between the authenticity score of the driving simulator and the difference between drivers and their driving behaviors.The differences between drivers’ subjective cognition and real driving behavior were analyzed by paired sample t test,and the coincidence degree was compared.It was found that the driver showed more dangerous driving behaviors than his subjective consciousness during the experiment,which mainly reflects on the frequency of keeping slow speed in side-parking road and the way to avoid collisions.Moreover,many suggestions for improvement of the warning technology were put forward,which provided new ideas for future research.