Characteristics Analysis Of Lane-changing Behavior Based On High-resolution Data

Traffic flow theory is the basis of ITS(Intelligent Transport System).As a kind of common drive behavior,lane-changing behavior is already a central issue of theoretical research.Limited to experimental technical conditions,on the one hand,post research results are usually based on low-resolution samples or data from simulation,on another hand,most existed research focus on single lane-changing but continuous lane-changing.Based on high-resolution data and the definition of lane-changing process,this paper proposes a new algorithm to extract single lane-changing and continuous lane-changing trajectory from original data,then analysis the character of lane-changing process according to the data extracted by the new algorithm.Firstly,this paper introduces the source of high-resolution data,choose the suitable pretreatment method of original data,analysis the relationship between definition and data extraction method,propose a new mixed search algorithm for single lane-changing start point and end point on the basis of lateral position and lateral speed,verify the superiority of new algorithm in accuracy and robustness.Secondly,this paper specify the main object of study by defining the type of lane-changing,calculate the minimum sample size according to statistical principle,select the method to handle abnormal data.Through in-depth research on lane-changing duration and longitudinal displacement,we find that existed research underestimate the duration of lane-changing which influenced by traffic density and cross-line time is the most affected stage in the three lane-changing stage while they are in low correlation.Through intensive study on the influence factor of the three lane-changing stage duration,we find that the significant influencing factors is different for the three stage of lane-changing process.In terms of longitudinal displacement,we find that the ratio of longitudinal displacement in trajectory divided by traffic line is bigger than 1.0.Finally,on the basis of above results,we further propose a search algorithm for continuous lane-changing process with the definition of lane-changing waiting duration.We find several feature about lane-changing waiting duration:it present as"L" shape while in 10s statistical scale and it shows slight increase at approximate 2s while in 5s statistical scale.In addition we propose a conception of lane-changing waiting threshold to deconstruct continuous lane-changing.We compare the net space between lane-changing vehicle and following vehicle in target lane,find that the net space and lane-changing interval at continuous lane-changing stage is bigger than it at single lane-changing stage according to 15%-quantile?85%-quantile and mean.Besides,both of them obey logarithmic normal distribution...