Study On The Gap Acceptance Behavior Under The Countdown Signal

As a common traffic management facility,the effect of countdown signals on road traffic,especially at intersections,has been a research hotspot in the field of road traffic safety.Up to now,many domestic and foreign scholars have studied the impact of countdown signals on non-motorized traffic,most of which are based on the change of pedestrian violation rate.However,the rate of violations does not fully reflect the changes in traversing behavior of traversers.At the same time,most studies ignore the problem of heterogeneity between individuals.Considering the above problems,this article,combined with the gap acceptance theory,innovatively explored the impact of the countdown signal on pedestrians and non-motor vehicles from the perspective of gap acceptance based on the data collected by field research.In addition,a random coefficients Logit model was established to solve the problem of heterogeneity among different individuals.Finally,prediction models of gap acceptance behavior are established based on the machine learning algorithm.The main work and research results of the paper are as follows:(1)Characteristic analysis of crossing gap under countdown signal.Based on the field survey data of three typical intersections in Beijing,this article fits the accepted / rejected gap values by use of non-linear fitting method.Analyzing the various factors' impact on the gap values by the multiple linear regression model of the crossing gap.Analyzing the change of the accepted gap in different crossing groups and different traffic environments with the help of hypothesis testing.The results show that the log-normal function has a better fitting on the distribution of crossing gaps;the factors that have a significant influence on the crossing gap value under the countdown signal are age,gender,waiting position,crossing phase,and the number of people crossing together;the countdown signal has a significant impact on the accepted gap between middle-aged people and women,and also has the same results during the red signal period and when the flow is large.(2)Modeling and analysis of gap accepted behavior under the countdown signal.By establishing a random coefficient logit model of gap acceptance behavior,the relationship between various influencing factors and the gap acceptance behavior of traversers is analyzed.The research results show that,compared with the traditional binomial logit model,the random coefficient logit model has a better goodness of fit.Factors such as crossing phase,waiting time,traffic volume,gap value and gap type have a significant impact on the gap acceptance behavior of the traverser.Among them,the impact of the waiting time on the decision of the traverser varies from individual to individual.(3)Gap acceptance behavior prediction under countdown signal.After using the Relief algorithm to filter out the variables entered into the model,the gap acceptance behavior prediction model based on C4.5 decision tree,BP neural network and SVM is established respectively.The results show that the 10 most important variables that affect gap acceptance behavior are gap value,waiting time,gap type,age,crossing phase,signal type,traffic volume,scale,waiting position and gender;compared with the other two models,the C4.5 decision tree model has better prediction performance,and its accuracy and AUC values are 86.65% and 0.90,respectively.