Intersection Traffic State Discrimination And Control Optimization Based On Vehicle Trajectory Data

With the proliferation of motor vehicles in recent years,traffic congestion has become a prominent manifestation of urban transport problems.Traffic congestion causes vehicles to travel only at low speeds,while increasing delays and exhaust emissions.As a key node in the urban road network,the operational status of the intersection is directly related to the operational efficiency of the road network.Scientific and rational traffic state identification and control optimization at intersections can help to reduce traffic congestion at intersections and reduce motor vehicle emissions,which is of vital importance for improving air quality and achieving "carbon peaking and carbon neutrality".Based on the vehicle trajectory data collected by wide area radar,this thesis uses theories and methods such as cluster analysis,extreme learning machine and population intelligence optimization algorithms to study the discriminative method of intersection traffic operation status,proposes an improved shorttime traffic flow prediction method and realizes intersection signal control optimization based on the short-time traffic flow prediction data.The main work of this thesis includes the following four areas.(1)Intersection vehicle trajectory reproduction based on data from wide area radar detections.The thesis analyses intersection vehicle data obtained by the wide area radar detectors and finds problems with the data such as detection anomalies,duplicate detection and duplicate numbering.The problems in the original data are resolved by processing the data through data cleaning,lane numbering and number resetting.Based on the unified coordinates of the relative positions between detectors,an intersection simulation platform is established to reproduce the process of intersection vehicle movements.Evaluation indicators such as traffic flow,average speed,queue length,vehicle delays and lane space occupancy are output to provide data support for the following study.(2)Intersection traffic state discrimination based on fuzzy clustering analysis.Three indicators,namely traffic flow,lane space occupancy and queue length,are selected to classify the vehicle operation status of the intersection into four classes using the fuzzy clustering algorithm.The experiment is conducted using the historical data of the intersection,and the results show that the affiliation values of the various traffic states obtained by the algorithm are mostly close to or equal to 1,indicating that the clustering effect of the algorithm is good and can effectively classify the traffic states of the intersection.(3)Short-time traffic flow prediction method based on improved ELM.The thesis proposes an improved extreme learning machine algorithm for short-time traffic flow prediction using an ant colony algorithm.The parameters of the extreme learning machine are optimized through ant colony algorithm search to improve the stability and prediction accuracy of the algorithm network.Experiments using real data show that the algorithm proposed in this thesis maintains good prediction results despite sudden changes in the data,verifying the stability of the algorithm.In terms of running speed,the algorithm proposed in this thesis also has a significant advantage,being 104 times faster than BP neural networks and 24 times faster than PSO-ELM.(4)Intersection signal optimization based on genetic algorithms.Based on the shorttime traffic flow prediction data,the thesis constructs an intersection signal optimization model and solves it using the genetic algorithm.Based on the intersection traffic state discrimination results,data from two time periods,representing two traffic states respectively,are selected to validate the model.The experimental results show that the optimized signal timing scheme reduces the average delay at intersections by 36.20% under congested conditions compared to the status quo;in the steady condition,the average delay at the intersection is better than the status quo by 37.67%.