| With the development of modern society,the problem of urban traffic congestion is increasingly serious,and the research on the coordination and optimization control of road network traffic has become a very important research field,which is of great significance to alleviate the traffic congestion of urban road network.Facing the complex urban road network and the changing traffic flow environment,the traditional methods have problems such as poor optimization and control effect.In view of the coordination and optimization control of regional traffic network,this paper studies based on sub-area dynamic division and state assessment.The main research contents are as follows:(1)Aiming at the problem of sub-area division in complex transportation networks,a dynamic sub-area division method is proposed with the goal of optimizing coordinated control,which combines the improved Louvain algorithm and the improved correlation degree model.Firstly,the traditional Whitson model is optimized by introducing the dispersion coefficient of traffic flow and the periodic coefficient of adjacent intersections.On this basis,an improved correlation degree model is proposed by comprehensively considering factors such as road attributes and traffic flow density.Based on the size of modularity as the criterion,the Louvain algorithm is optimized by improving the criteria for determining the increment of modularity,making it more effective in dynamically dividing traffic subregions.Finally,select the actual regional road network for comparative and validation experiments.The experimental results show that the proposed improved model partitioning method can better integrate actual traffic flow characteristics,achieve more reasonable real-time dynamic partitioning of sub-area in the regional road network,and provide a good foundation for coordinated control of subsequent sub-area.(2)On how to effectively evaluate the status of traffic sub-areas and classify the ranking,on the basis of the sub-area division is completed,A state assessment method of matter-element extension model with improved index weight allocation mechanism is proposed.Firstly,a matter-element extension model is introduced,which combines complex network parameter indicators and traffic characteristic indicators,improve the allocation mechanism for assigning weights to indicators.Set up three experimental plans,the main variable of the experiment is the method of determining the index weights,they are principal component analysis method,optimization entropy weight method,and comprehensive improvement weight method.Compare and evaluate the evaluation plans for the state of the three seed zones through experimental results.The experiment has proven that the proposed sub-area state evaluation method is more effective and reasonable.The evaluation results can clearly rank and sort each subregion;combined with the visual coordinate image,the priority order of traffic optimization is also more directly displayed.(3)Aiming at the problem of long optimization time during the traffic control in sub-areas,introduce deep reinforcement learning,propose a sub-area coordinated control strategy that integrates the improved discount coefficient DQN algorithm.Based on the traditional DQN algorithm,the time error of the agent in the multi-agent system during the action execution is comprehensively considered,that is the time error of the execution of the action after feedback exists between the sub-area.By improving the discount factor in the reward function,the sub-area are more inclined to maintain the same green light duration,the improvement of the discount coefficient also optimizes the performance of the DQN algorithm,thereby better achieving signal coordination optimization between various traffic subregions.The experiments show that The improved DQN algorithm has a more stable performance and better for optimized control of sub-areas coordination.Through three benefit evaluation indexes of vehicle delay,queuing length and parking number,we prove that the proposed sub-area coordination control scheme can alleviate the traffic congestion of the regional road network more effectively. |
