| As the urban motorization was accelerated,the prominent problem of urban traffic congestion can be seen.A key problem causing traffic congestion was reflected in the lane-changing of vehicles during the morning and evening peak of urban sections.Due to the random or forced vehicle lane-changing,the traffic environment will become worse,which leads to the deceleration or stop of the rear vehicles in the target lane,and causes the traffic accidents and other risks in serious cases.In this thesis,the high-density traffic flow(0.80?V/C?0.90)of expressway(South Second Ring Expressway of Xi'an City)was taken as the research background and vehicle lane-changing as the research content,the process of vehicle lane-changing and the impact of lane-changing on the rear vehicles in the target lane are discussed.Firstly,the extraction of lane-changing data was carried out,the UAV was adopted to vertically got the video of the expressway section,the Tracker software was used to extract vehicle information,and the LOWESS smoothing method was used to smooth the extracted data.In line with the survey data,it started from the perspective of system to analyze the speed,insertion angle and vehicle spacing of lane-changing vehicles,and the construction of Bayesian network lane-changing model was made.The Bayesian network lane-changing model was used for machine learning by adopting MMHC algorithm and MMHC algorithm based on I-CH scoring criterion,in which it was found that the improved I-CH scoring MMHC algorithm had better effect.The appllication of the Asia network was made to test the lane-changing model,so the prediction accuracy of the vehicle lane change model was obtained as 86.41%.After conducting the Bayesian network learning of vehicle insertion angle,speed and spacing,it was concluded that the vehicle can be judged to change lanes when the changing rate per unit time at the initial moment of vehicle lane-changing insertion angle was greater than 3.0°.In the meantime,it was also concluded as follows,the vehicle lane-changing success rate was the highest if the speed of first rear vehicle in the target lane was in the range of 1.0-3.0m/s,and the front-rear vehicle safety spacing was less than 4.0m;the success rate of vehicle lane-changing was the highest when the front-rear vehicle safety spacing was greater than 4.0m,and the speed of first rear vehicle in the target lane was in the range of 3.0-6.0m/s.Secondly,the construction of a Gaussian Mixture Model Hidden Markov Delay Model was carried out in line with the relationship characteristic between the target vehicle lane-changing process and the rear vehicles in the target lane.The modeling process was divided into two stages.The impact of the target vehicle lane-changing on the first rear vehicle in the target lane was discussed in the first stage,and the main factors considered include the ordinate and speed of the target vehicle.The influence of the first rear vehicle on the second rear vehicle and the influence of the rear vehicle n on the rear vehicle n+1 in the target lane were studied in the second stage,in which the main factors to be considered include vehicle speed,distance and so on.From the above two stages,the construction of the model was conducted.The prediction was that there were 81.20%,66.71%,52.24%,27.36%and 10.95%of the time in the lane change time(8) for first,second,third,fourth,fifth rear vehicle to decelerate.In the meantime,it got the conclusion that the lane-changing of the target vehicle will cause 50%of the delay of 5.0-10.0s in the target lane,28%of the delay of more than10.0s,22%of the delay of less than 5.0s.The average absolute error of the model was12.81%.Finally,the establishment of the vehicle lane-changing risk assessment model was conducted,in line with the vehicle lane-changing probability and the vehicle lane-changing consequences,to predict the evaluate vehicle lane-changing risk of the rear vehicles in the target lane,and the four-level risk of vehicle lane-changing was divided.Level 1 risk was the largest with the risk value range of[2.091,+?;Level 2 risk was relatively large with the risk value range of[1.404,2.091);Level 3 risk was general with the risk value range of[0.717,1.404);Level 4 risk was low with the risk value range of[0.03,0.717).A vehicle lane-changing risk early warning system was constructed from the perspective of the system,in which the four modules such as data acquisition module,lane-changing recognition module,lane-changing risk assessment module and early warning module were included.When carrying out lane-changing early warning,it got the conclusion that the insertion angle and speed of the target vehicle were the key factors that generate the risk of vehicle lane-changing according to the entropy method.By starting from the key risk factors,in the expressway traffic load of 0.80--0.90 state,it was recommended that the insertion angle should be kept the range of 15.0°--21.0°as much as possible,and the vehicle speed should be kept below 5.0m/s when the vehicle was changing lanes,and the risk of the vehicle lane-changing to the rear vehicles was relatively small.This paper was besed on the research of the vehicle lane change behavior under the state of high-density traffic flow,and provided methods for analyzing the risk of vehicle lane change.The research results provided theoretical support for vehicle movement control and policy formulation of urban expressways. |
PDF Full Text Request