Modeling Analysis Of Differential Lane-Changing Behavior In The Open-Section Interweaving Area Of Urban Continuous Tunnel

The interweaving area of open sections of urban continuous tunnels is a high-incidence area of urban traffic problems.A large number of vehicles changing lanes are prone to traffic congestion,traffic accidents,etc.,which affect the normal traffic order.The purpose of studying the lane changing behavior in this area is to improve the overall traffic operation efficiency and reduce the occurrence of traffic accidents.Most of the current research analyzes the lanechanging process as a whole.In this study,the lane-changing process is considered to be divided into two stages: lane-changing decision-making and lane-changing execution.Analyze differential lane-changing behavior during the lane-changing process,and construct a lanechanging decision-making model based on the differentiated lane-changing behavior,provide a theoretical basis for the formulation of traffic safety management plans and the establishment of future automatic assisted driving systems.Firstly,collect experimental data through aerial photography,and combine the characteristics of lane-changing behavior to establish the influencing factors of vehicle lanechanging behavior from three aspects: vehicle individual driving characteristics,traffic flow environment characteristics and lane-changing execution characteristics.Two lane-changing behavior datasets are formed through data preprocessing,lane-changing decision and lanechanging execution.Secondly,according to the lane-changing purpose of the lane-changing vehicle and the area where the lane-changing decision point is located,the lane-changing decision-making environment is differentiated.The random forest is used to analyze the feature importance,and the support vector machine is selected to establish the lane-changing decisionmaking model.The decision-making situation is reflected by the lane-changing decision probability.Then,based on the lane-changing behavior target lane lead distance and lag distance,the hierarchical clustering method is used to differentiate the lane-changing execution behavior environment(6 categories),and the deep neural network regression is used to analyze the lanechanging execution process in different environments.model.Finally,the data of the interleaving area in different time periods are selected to verify the model,and the lane-changing decision model is used to predict the lane-changing probability.It has been verified that among the 40lane-changing probability values,only 4 have an error ratio of more than 50%,and the lanechanging model has an accuracy rate higher than 75% in describing behavior.Under the environment of 6 types of different lane-changing execution behaviors,the model’s prediction values of the lane-changing duration and lane-changing duration are significantly improved compared with those without environmental classification,and the prediction accuracy of the lane-changing duration under the classification environment is improved by at least 61%.Above,the lane change duration prediction accuracy is improved by more than 57%.In this study,lane-changing behavior is divided into different stages,and based on traffic flow data,a lane-changing decision model based on random forest and support vector machine and a lane-changing execution behavior model based on deep neural network are constructed.Segment lane change behavior process.The research results can be used to improve the lane change decision assistance system and the establishment of microscopic traffic simulation models,thereby improving the safety and effectiveness of the vehicle automatic driving assistance system,and improving the authenticity and accuracy of the simulation output results.