Vehicle Speed Guidance And Traffic Signal Cooperative Optimal Control In Urban Road Network

As an important node in the urban road network where traffic flows converge,turn and diverge,the signalized intersection is the key to ensuring smooth urban traffic flow.The traditional intersection control methods can not effectively deal with complex traffic environments with more spatio-temporal conflicts.Against this background,intelligent vehicle-infrastructure communication technology realizes real-time dynamic information interaction between vehicles and infrastructures in a wide range and in an all-around way,which provides a technical basis for exploring new means of intersection traffic control.On the basis of making full use of the controllability of connected vehicles and advantages of vehicle-infrastructure communication,the speed guidance and traffic signal control under multi-level traffic environment are systematically studied.The main work includes the following aspects:(1)Study on micro modeling of mixed traffic flow of connected/non-connected vehicles.With regard to the respective driving characteristics of the two types of vehicles approaching the intersection,a vehicle motion model and a driver behavior model considering individual characteristics are integrated and designed,and a human-vehicle system model is established with the combining of "mechanistic analysis" and " data-driven".Taking advantage of multisource data of intelligent vehicle-infrastructure communication,the model is trained based on vehicle data such as vehicle spacing,relative speed of vehicles,relative distance to intersections,signal phase and countdown,gas pedal and brake pedal,and speed guidance information,by using neural network and Gaussian Mixed Hidden Markov Model.Vehicle trajectory prediction considering individual vehicle characteristics,and dynamic prediction of intersection congestion and delay are realized to provide a model basis for intelligent traffic control considering mixed traffic flow environment.(2)Research on speed guidance and cooperative control of speed guidance and traffic signal under mixed traffic environment of connected/non-connected vehicles.A dual-objective collaborative optimization control model of vehicle speed guidance and traffic signal control is established based on vehicle trajectory prediction at intersections.According to real-time vehicle state and signal phase and timing,optimal speed guidance is carried out for the mixed group composed of connected/non-connected vehicles.Meanwhile,the dual-objective cooperative optimization control model for single intersections is extended to the multi-stage dual-objective cooperative optimization control model for multi-intersections in the arterial road.By distinguishing light/heavy vehicles in the speed guidance and adopting a hierarchical approximation solution framework,the problem dimension and solving complexity are reduced,the cooperative speed guidance and traffic signal control in the regional road network is finally realized.(3)Study on traffic control problem of complex intersections with tram and connected/non-connected vehicles.Based on the real-time dynamic information interaction between vehicles and infrastructures,a novel tram signal priority optimization control model based on sequential decision theory is proposed.By processing high-dimensional traffic data through deep neural networks,and then effectively obtaining the real-time dynamic changes of connected/non-connected general vehicles,the signal priority of trams under the premise of autonomous speed optimization of connected vehicles is realized.While ensuring the priority of tram,the congestion of connected/non-connected common vehicles is reduced effectively.In view of the complexity of the problem,a control strategy design method by integrating model-driven and data-driven methods is proposed.Combining the advantages of model predictive control and deep reinforcement learning,a multi-step signal optimization control algorithm for tram has been constructed which can effectively deal with real-time traffic changes.(4)Research on cooperative control of large scale road network with tram,bus and connected/non-connected general vehicles.A multi-mode traffic signal control model based on multiagent deep reinforcement learning is designed to address the limitation of centralized optimization control.A unified quantization method based on the number of passengers is established to integrate the real-time dynamic information of all kinds of traffic vehicles.In order to alleviate traffic congestion in road network and achieve the optimal performance of the joint multi-mode traffic,the single-agent model-driven approach and multi-agent data-driven approach are been combined to first realize the local intersection-based control with independent intersection model prediction control,and then realize the collaboration among regional multi-intersections with multi-agent reinforcement learning approach,and finally establish the joint multi-mode traffic control algorithm of large scale road network.