Study On Consecutive Conflicts Of Pedestrian Crossing Based On Binary Logit Model

With the explosive growth of the number of non-motorized vehicles and vehicles in China,the crossing environment of pedestrians is increasingly complex.At the uncontrolled traffic environment,pedestrians often have traffic conflicts between pedestrian-vehicles and pedestrian-non-motorized vehicles that occur consecutively during the process of pedestrian crossing.The pedestrian crossing is becoming more dangerous,and it is necessary to study on the influencing factors of pedestrian crossing safety.Considering that it is difficult to collect related accidents,traffic conflict technology is adopted to replace the statistical analysis of traffic accidents to study pedestrian crossing safety.Firstly,the definition and discriminant method of consecutive conflict are defined,and the discriminant basis to distinguish continuous conflict from multiple conflicts is proposed.In order to quantify the severity of pedestrian crossing,considering the one-sided problem of one-dimensional indicators,combining with spatio-temporal proximity index PET and hedging behavior index YRR and Jerk,a unified indicator STC of conflict severity is constructed by mapping function,and the definition standard of conflict severity is given.A total of 9 hours of videos are recorded by UAV at 6 non-signalling roundabouts of an interchange in Shenyang.T-Analyst is used to process the recorded videos,establish a track database,and output indicators such as PET,YRR,Jerk and STC for future research.Secondly,the conflict discrimination of 925 groups of consecutive conflicts(1850 single conflicts)is carried out,and the discriminant results of single dimension index and joint measurement index are compared.Then,according to the number of consecutive conflicts and the information of non-motorized vehicle and vehicle flow,the law of conflict generation is analyzed,and the prediction model of consecutive conflicts is built on the assumption that consecutive conflicts obey the negative binomial distribution.In this paper,STC value is used as as the basic criteria conflict order safety,conflict object security,site security,the results show that: For a consecutive conflict,the severities of two individual conflicts are different,and the difference is not related to the sequence of conflict but to the types of conflict subjects.The conflicts between pedestrians and non-motorized vehicles are in general more serious.The risks of pedestrian crossing are higher at converging points than those of diverging points.Thirdly,based on the behavioral characteristics of various road users and related studies,10 variables related to pedestrian crossing conflicts are identified,and the multicollinearity among the variables is tested.Two unary Logit models for predicting the severity of the conflict are constructed,with the severity of the conflict(serious =1,non-serious=0)as the explained variable.At the same time,considering the correlation between two consecutive conflicts,the binary Logit conflict severity prediction model is constructed under the same modeling framework.Based on Bayesian estimation and Markov chain Monte Carlo(MCMC)simulation method,the parameters of three conflict severity prediction models are calibrated,the correlation between pedestrian-non-motorized vehicles conflict and pedestrian-vehicles conflict is analyzed,and DIC value is used to compare the models,the comparison between the results show that: the binary Logit conflict severity prediction model of consecutive conflicts is superior.Finally,the OR value is introduced to qualitatively and quantitatively analyze the correlation between each influencing factor and conflict severity,and the causes and mechanisms are analyzed in combination with the correlation.According to the quantitative analysis,the factors influencing the severity of conflict are divided into four aspects: speed(v＿pedes,v＿cyc,v＿veh),number of vehicles(n＿cyc,n＿veh),safety margin(ttz1)and crossing behavior(rolling and direct),and corresponding preventive measures or improvement suggestions are put forward according to four aspects.Through the identification and modeling of conflict of pedestrian crossing,this study puts forward a comprehensive conflict severity measure method,and establishes a binary logit model that can reveal the correlation of consecutive conflicts occurring during pedestrian crossing.The findings can enrich and perfect the traffic conflict technology system and pedestrian crossing safety analysis method.In addition,based on the results of modeling analysis,this paper puts forward policies and measures to improve pedestrian crossing safety in no-signal environment,which also has certain practical guiding value.