Research On Dynamic Coordinated Control Method Of Traffic Signals Based On Historical Trajectory And Real-time Data

Currently,the number of automobiles in use is growing rapidly,and the demand for urban road traffic is approaching or exceeding the supply of road capacity,especially in morning and evening rush hour scenarios.This has resulted in frequent traffic congestion issues.Relying solely on newly added road infrastructure is no longer sufficient to meet transportation needs.There is an urgent need to propose new methods for traffic management.It is necessary to fully utilize the existing electric police checkpoints and signal control facilities at urban road intersections;And it is necessary to fully utilize the limited time and space resources of urban roads;Furthermore,more precise and efficient traffic signal control methods can be adopted.To ensure efficient operation of vehicles on the road and alleviate urban traffic congestion.Therefore,this study analyzes the checkpoint data,and divides it into vehicle data with or without historical trajectories for traffic flow prediction and stacking.A traffic flow prediction model based on the fusion of historical trajectories and real-time data has been established to obtain more accurate traffic flow data.Based on this as the fundamental data for traffic signal control,a multi-objective optimization based adaptive control method for single intersections and dynamic coordination of mainline signals was constructed.The specific research content is summarized as follows:Firstly,the research background and significance of the article were analyzed.The current research status of vehicle trajectory mining,short-term traffic flow prediction,and traffic signal control were reviewed both domestically and abroad.The existing problems and deficiencies were analyzed.And the main research content and structural framework of this article were introduced.Secondly,data collection and analysis are the main content of this section.The checkpoint data were extracted,and the analysis and preprocessing of abnormal data were carried out.Methods such as data cleaning,data screening,similarity detection,data matching,and missing data repair were used to correct erroneous data,remove abnormal data,and supplement missing data.Thirdly,the individual vehicle identification and trajectory extraction were carried out using the license plate recognition technology of the road intersection checkpoint system and the database system of the traffic management department.A time threshold was introduced to define the segmentation point of the vehicle’s daily trajectory,to divide the vehicle’s single travel trajectory.And a historical vehicle trajectory database was established.And the macro and micro operational rules of the vehicle’s historical trajectories were analyzed to establish the basis for traffic flow prediction.Fourthly,using the Stacking principle,a short-term traffic flow prediction model based on ensemble learning was constructed.The basic learners selects three models: SVR,LSTM,and XGBOOST models.And the second-level learner used the ridge regression model for model fusion to prevent overfitting.It has been verified that the traffic flow prediction model based on ensemble learning fusion is superior to a single prediction model.This article divides the real-time collected tollgate data into two categories.One category is checkpoint data with historical vehicle trajectory information.Based on the rules of historical vehicle trajectories.This part of data can accurately predict the driving direction of vehicles at the next intersection.The other category is checkpoint data without historical vehicle trajectory information.By using a traffic flow prediction method based on ensemble learning fusion,this part of data can be used to predict the traffic flow volume.Based on the distribution pattern of historical traffic flow in different directions,weight coefficients were determined to allocate and predict the traffic flow in different directions(straight,left turn,right turn)of the intersection entrance.Then,these two types of data were superimposed to calculate and predict the traffic flow in different directions of the signalized intersection with high accuracy.Taking the north entrance of Liuquan Road to Zhongrun Avenue during the evening peak period as the experimental object,it was found that the Stacking fusion model had better predictive performance than a single traffic flow prediction model;A traffic flow prediction model that considers the fusion of historical trajectories and real-time data can further improve the prediction performance of the Stacking fusion model.Finally,based on dynamic prediction of traffic flow in various directions at intersections,a multi-objective optimization based adaptive signal control method for signal intersections under different saturation levels(unsaturated,saturated,and oversaturated)were proposed.By selecting different optimization objectives and searching for solutions through genetic algorithms,the dynamic optimization of intersection signal control schemes can be achieved.The VISSIM platform is used for microscopic traffic simulation to verify the effectiveness of dynamic coordinated control method.In the simulation results,the single point adaptive control scheme has better performance than the current control scheme,with a decrease of8.82% in vehicle delay,a decrease of 7.21% in average parking times,and a decrease of12.84% in average parking time;The maximum green wave band method is used in the flat peak phase of mainline signal coordination control.Compared with the current situation,this method reduces average number of stops by 11.29%,and reduces average vehicle delay by16.45%.Compared with the traditional Webster timing method and the current situation,the multi-objective optimization algorithm used in the peak stage has the best control effect.Among them,compared with the current average number of stops,the use of Webster method reduced by 6.56%,and the use of multi-objective optimization reduced by 20.98%.Compared with the current average vehicle delay,the use of Webster method reduced it by8.27%,and the use of multi-objective optimization reduced it by 16.90%.In summary,during the flat and peak periods,a dynamic coordinated control method for urban traffic signal control based on historical trajectory data and real-time data can achieve dynamic allocation and optimization of road rights at intersections.It can efficiently utilize spatiotemporal resources of intersections,effectively reduce the average number of stops and vehicle delays at intersections and improve road traffic efficiency.