Research On Lane-changing Event Discrimination Method Based On Driving Trajectory

In recent years,with the increase of car ownership,traffic congestion has become more and more serious.Lane-changing event,as one of the events that occur easily in the process of vehicle driving,will not only affect the operation efficiency of road traffic flow to a certain extent,but also cause traffic conflicts and affect road traffic safety.Therefore,it is one of the hot topics in the field of intelligent transportation to study the lane-changing event and how to make the following vehicle predict the lane-changing event quickly and effectively.At present,most of the lane-changing state of auxiliary system through a turn signal to determine whether lane-changing events will happen,but in the actual incident vehicles turn signal in advance in the process of opening rate is low,the system can’t change the way events timely and accurately judging,if can be combined with other lane-changing feature index to identify whether lane-changing event occurs,not only on the basis of this single turn signal indicators,The driving environment around the driver can be evaluated in advance to reduce potential driving risks.In this paper,based on a large number of vehicle trajectory data,vehicle speed,lateral acceleration,lane-changing time,turn signal on,relative distance,collision time and so on were selected as the characterization parameters of lane-changing events.The Logistic model was introduced to distinguish lane-changing events,and the discriminant method of lane-changing events was proposed.The main work of this paper is as follows:(1)Select some road sections with good traffic flow in Suzhou City,use the video shooting method to collect the vehicle trajectory data,use the Tracker data processing software to extract the vehicle trajectory from the video,and convert the trajectory coordinates based on the camera imaging principle.The subsequent driving trajectory is smoothed,the starting point of the lane-changing is determined by the change of the vehicle’s lateral position,and the data is screened into lane-changing event samples and lane keeping event samples to lay the foundation for subsequent analysis of lane-changing characteristic indicators.(2)The whole process of lane-changing event was extracted from the driving track data,and the variation rules of vehicle operating parameters in the lane-changing stage and lane-keeping stage were studied in depth.The results showed that;There were significant differences in each parameter.About 68% of the lane-changing samples had a speed range between 45 km/h and 60 km/h.The mean standard deviations of lateral acceleration of lane-keeping and lane-changing were 0.21m/s^2 and 0.07 m/s^2,respectively,and the differences were significant.In the lane-changing samples,the turn signal on rate was about 78.5%,and only 45.3% of the samples were turned on by the beginning of lane-changing.The general trend of lane-changing time distribution is relatively concentrated,concentrated between 3-6s.At the beginning of lane-changing,the relative distance 1 between the self-driving vehicle and the vehicle in front of the current lane concentrates between 50-60 m,and the collision time 1 is about 10 s.The relative distance 2between the autonomous vehicle and the vehicle behind the target lane is concentrated at about 50 m.When relative distance 2 is greater than 50 m,the collision time 2 is about 15 s.When relative distance 2 is less than 50 m,the value of the collision time 2 is more than20 s.(3)Based on the analysis of the difference of related parameters,the indicator system for identifying lane-changing events is determined by vehicle speed,lateral acceleration,turn signal on,vehicle relative distance and collision time,and a lane-changing event identification model based on Logistic regression is constructed.And using the samples to be identified to test the prediction accuracy of the model,the overall accuracy rate reached93.2%,indicating that the model has a good prediction effect,and the method of identifying lane-changing events through indicators and models is effective.