Research On Discrimination And Prediction Methods Between Vehicle And Bicycle Conflict At Signal Intersection

With the development of bike-sharing such as OFO,Mobike and Bluegogo,traveling with bicycles becomes more popular and plays an important role in city traffic as well as daily life.Signalized intersection,as an important node of city roads,directly reflect the operation efficiency of the road network.The traffic flow at the intersection is complex which causes traffic conflict frequently.The traffic collision between the right-hand vehicles with the other traffic flow at the intersection seriously affects the traffic efficiency due to the lack of right-turn phase at the most of the intersection in our country,especially the traffic conflict with the bicycles.This kind of conflicts not only causes delays and affects the traffic efficiency of the right-turning vehicles,but also poses threats directly to the safety of the bicycles crossing the street which is the main reason of traffic accidents.The paper focuses on the conflict between right turn vehicles and bicycles at signal intersections,summarizes the domestic and foreign researches,including traffic conflict,data acquisition,identification,safety evaluation and applications.In order to reduce the conflicts between the right turn vehicles and the bicycle,improve the safety of the bicycles and the safety of intersections,this paper includes the following four aspects:(1)Studying the traffic characteristics of different types of traffic flow at the signalized intersections,especially the traffic characteristics of right-turning vehicles and bicycles.The conflict characteristics of right-turning vehicles and bicycles includes conflict area,distribution of conflict points and so on.(2)Using artificial survey method and video acquisition method to study the typical intersection,combine with the video software and data extraction software to extract and analyze the conflict data of the intersection,which can be used to provide large amount of data for the identification and prediction of traffic conflicts.(3)Using the conflict time and conflict distance as the discriminant indexes respectively.K-Means clustering analysis is used to determine the conflict and the severity of obtained data.(4)Analyzing the factors which influences the conflict,using multiple linear regression and BP neural network respectively to build the prediction model of traffic conflict,analyzing the prediction of the two models.Compare the prediction of the two models,support the actual conflict prediction finally.