Research On Dynamic Coordinated Control Of Urban Arterial Traffic

With the increasing number of urban motor vehicles in our country,the urban traffic congestion has become one of the main factors restricting the current urbanization process.The urban trunk road is the artery of the urban road network.The research on the coordinated processing of signal control between the intersections of the urban trunk road is of great significance to alleviate urban traffic congestion.In this dissertation,the dynamic coordinated control is researched to reduce the average delay and the number of stops of the urban trunk road.The main work of the dissertation is as follows:1.Deep Q network method for dynamic coordinated control of urban arterial trafficThis dissertation introduces the deep Q network algorithm into the dynamic coordinated control of urban arterial traffic,and treats all intersections of the urban trunk road as a whole.This method extracts the traffic state characteristics of the trunk road through the deep neural network,finds the correlation of traffic signal control at multi-intersections,and uses Q learning algorithms to make traffic signal control decisions.In this dissertation,the state space,action space,reward function,and deep neural network structure related to the dynamic coordinated control are designed in detail in accordance with the traffic characteristics of the trunk road.The effectiveness of the dynamic coordinated control of urban arterial traffic based on the deep Q network algorithm design was verified by the simulation experiment platform.2.Dynamic coordinated optimization control of urban arterial traffic based on improved deep Q networkIn order to further improve the effect of dynamic coordinated control of the trunk road,this dissertation integrates the double deep Q network and the dueling deep Q network,and designs dynamic coordinated optimization control based on DDDQN(Dueling Double Deep Q Network).The over estimation problem of the deep Q network and the structure of the deep neural network are improved at the same time,and the effects of random sampling and priority experience replay on the arterial coordinated control are compared.The results of simulation experiments show that:DDDQN-based arterial coordinated control has the best effect,and the use of random sampling for exploratory learning can better reduce the correlation between sampled samples,and it performs better in the near saturation and over saturation.3.Implementation of parallel transportation system for arterial coordinated controlDue to the high risk of experiments in the actual traffic environment,this dissertation introduces the parallel transportation system method to verify the effectiveness of the dynamic coordinated control method of the trunk road through the simulation experiment platform,so as to its practical application in arterial coordinated control provide support.Further,in the urban traffic signal intelligent control system,the arterial coordinated control module is designed and implemented.