Distributed And Hierarchical Optimal Control For Urban Traffic Networks

Traffic congestion in urban traffic networks becomes a serious problem all over the world.From a long-term perspective,traffic signal control on the basis of the available transportation infrastructures is a promising way to alleviate traffic jams.In this methodology,the traffic flow process is modeled by use of the macroscopic modeling approach,and the problem of traffic signal control including all constraints can be formulated.If the corresponding optimization algorithm is embedded in a rolling-horizon control scheme,this method can be implemented in real time.However,urban traffic networks are large-scale systems,consisting of many intersections controlled by traffic lights and interacting connected links.In practice,control of such large systems is often infeasible by only using a single controller,that is,in a centralized way.Therefore,in order to regulate the traffic flows and improve the mobility in cities,network-wide control of large-scale urban traffic networks using a distributed or multi-level hierarchical framework can be more efficient and flexible than centralized strategies for reducing the traffic congestion in big cities,because it can adequately address some problems that occur in controlling such large systems,e.g.computational complexity,multiple control objectives,weak robustness to uncertainties and so on.The main contributions of this dissertation are summarized as follows:· Investigation on network partition for large-scale urban traffic networks.In order to control the large-scale urban traffic networks by using distributed or hierarchical structure,it is necessary to exploit a network partition method,which should be both effective in extracting subnetworks and fast to compute.Based on consideration of both the physical characteristics and the dynamic traffic information of the link,the concept of correlation degree is proposed to determine the desire for interconnection between two adjacent intersections.It can be used as the weighted value of a link in an urban traffic network.Then,a fast network division approach by optimizing the modularity is applied to identify the subnetworks.Furthermore,considering the requirement of the MFD concept,we also present a method to decompose a heterogeneous urban traffic network into several homogeneous subnetworks based on the definition of similarity,which reflects the difference of traffic states between two adjacent links.· Investigation on distributed MPC control for large-scale urban traffic networks.Based on the decomposition of a large-scale urban traffic network,several agents are developed and allocated to the corresponding nonoverlapping subnetworks.Each agent employs a model-based predictive control approach and is able to make decisions by negotiating with its neighbors.In order to help the agents to reach an agreement on their decisions regarding traffic control actions as soon as possible,we propose a multi-agent control approach using a congestion-degree-based serial scheme to deal with the interactions among subnetworks.· Investigation on how to improve mobility in urban traffic networks by manipulating the traffic flows that transfer among subnetworks.A two-level hierarchical MPC control framework for coordinating the exchange flows among subnetworks is proposed that is capable of addressing different problems at different layers.At the upper level,based on decomposing a heterogeneous traffic network into several homogeneous subnetworks,a higher-level optimization problem using the concept of MFD is formulated to coordinate the exchange flows among subnetworks.At the lower level,the controller with a more detailed traffic flow model for each subnetwork determines the optimal signal timing within the given region under the guidance of the upper-level controller through communication.For the application of this architecture in real time,the MPC approach is utilized so as to obtain the best solution for both levels.Moreover,in order to decrease the computational complexity,a distributed control scheme within each subnetwork is developed at the lower level.· Investigation on how to improve mobility in urban traffic networks by regulating the traffic demands from outside.We concern the integration of demand balance control and traffic flow coordination control in a two-level hierarchical MPC control framework for complex urban traffic networks.At the first level,based on network partition,an improved MFD-based model is developed to describe the traffic dynamics in each subnetwork.This model aims at designing a demand balance MPC controller,which can improve the internal flow inside the subnetwork by regulating the input flows from outside.At the second level,based on a more detailed traffic flow model,the signal timings of all intersections for each subnetwork are determined by a flow coordination MPC controller,which aims at distributing the number of vehicles in each subnetwork as homogeneously as possible.