Font Size: a A A

Research On Regional Traffic Control Technology In The Network Environment

Posted on:2021-03-30Degree:MasterType:Thesis
Country:ChinaCandidate:Y C ZhangFull Text:PDF
GTID:2392330611480414Subject:Control engineering field
Abstract/Summary:PDF Full Text Request
In recent years,the continuous development of traffic signal control technology has effectively solved some urban traffic congestion problems.With the changes of travel mode,driving behavior and road infrastructure,the research of signal control technology has entered the stage of intelligent research.The extensive application of Internet technology and the rapid development of big data,5G,AI and other more accurate perception new technologies provide data and platform for the research of intelligent traffic signal control.Taking the intersection as an intelligent node and the real-time change of the traffic flow as the basis of the autonomous decision-making control of the traffic area is one of the urgent problems in the field of traffic signal control.Therefore,the research on the artificial intelligence signal control algorithm based on the agent technology has become a hot research direction.Based on the application of traffic data in the network environment,this paper establishes an agent model of signal control intersection which can interact with the environment,analyzes the game relationship between each intersection in the signal control area,constructs a coordinated control method of regional traffic signals based on multi-agent technology and artificial intelligence control algorithm,and finally improves the signal control of the whole road network System effect and operation state.The main work of this paper includes:First,research on the intelligent control method of single intersection.Intersection is the most basic control unit of signal control and the precondition of regional traffic coordinated control.In this paper,based on the interaction between the network data collected from the intersection and the signal controller,the control optimization algorithm is studied based on the deep reinforcement learning,and the signal control intersection agent is constructed.The multi index weight analysis method is used to construct the state space of the intersection agent,which makes the state space of the model more close to the actual traffic environment.Furthermore,the action phase,reward function and action selection strategy of the model are constructed.The drqn network model is used for training,and the memory ability of the circulating neural network LSTM is used to improve the optimization effect of the network model.The algorithm is designed by Python language and the VISSIM traffic simulation platform is built.The experimental results show that the queue length and vehicle delay of the intersection are optimized by 10.5% and 9.8% respectively.Secondly,research on the coordinated and optimized control technology of regional traffic intelligence.Static organization optimization is the most basic means to improve the regional traffic operation state.This paper studies the static organization optimization of Zhongguancun West area,introduces the game theory and dyna algorithm framework into the multi-agent structure model to study the regional traffic coordination control algorithm,divides the regional road network model into two levels,the upper level is the traffic sub area agent,the lower level is the signal control in the traffic sub area Intersection agent.Aiming at traffic subarea agent,a traffic subarea coordination control algorithm based on dyna-q is proposed by combining dyna algorithm framework and Q-learning.Aiming at the signal control intersection agent in the traffic sub area,the Stackelberg game is introduced to construct the internal control algorithm of the traffic sub area based on stackelberg-q.The synchro traffic simulation software is used to simulate the traffic organization optimization effect of the road network and the control effect of the regional traffic intelligence coordination optimization control algorithm.The experimental results show that compared with the traffic organization before optimization,the operation state of the road network is improved.Compared with the traditional large period fixed time allocation scheme based on expert system,the regional traffic intelligence proposed in this paper The coordinated optimal control algorithm improves the control effect of the road network slightly.
Keywords/Search Tags:Multi-agent technology, Deep Q-learning, Game theory, Regional coordinated control, Traffic organization optimization
PDF Full Text Request
Related items