Large-scale traffic signal control and multimodal network design

Traffic congestion constitutes one of the most frequent, yet challenging, problems to address in the urban space. Caused by the concentration of population, whose mobility needs surpass the serving capacity of urban networks, congestion cannot be resolved in the long term by creating more road space. Instead, network capacity can be increased by maximizing efficiency of traffic operations and road space use. Several methods can be employed for this purpose, including optimization of public transportation systems operations and intelligent, adaptive traffic signal control. This thesis contributes to the existing research in these directions, by developing and evaluating new modeling, optimization and adaptive signal control approaches, aiming at improving mobility in highly-congested, large-scale networks, while considering the dynamic characteristics of congestion propagation. The problem of optimal Dedicated Bus Lanes (DBL) location assignment in large networks with existing bus systems of fixed operational characteristics is addressed in chapter 2. A combinatorial optimization problem is formulated on the basis of an enhanced version of Store-and-Forward paradigm, a dynamic, queue-based macroscopic traffic model, able to properly capture the dynamics of backwards propagation of congestion due to queue spill-backs. Changes in mode choice equilibrium are considered in the evaluation of candidate solutions. An algorithmic scheme based on Local Search, problem-specific heuristics and Large Neighborhood Search (LNS) metaheuristic is developed to address the complex problem. Various destroy and repair operators for LNS are proposed, together with a learning process for assessing the importance of links in terms of receiving DBL, and a network decomposition strategy for accelerating the solution process for very large networks. A two-layer hierarchical traffic-responsive signal control framework is proposed in chapter 3, combining aggregated multi-region perimeter control (PC) with distributed Max Pressure (MP) control in isolated intersections. Partial deployment of MP in subsets of network nodes is performed and a methodology for identifying critical nodes for MP control based on node traffic characteristics is developed, in the scope of reducing MP implementation cost. Various node layouts of different network penetration rates are evaluated, both in independent MP application and as part of the combined framework. Simulation experiments are performed for a large-scale network of more than 1500 links and 900 intersections for two scenarios resulting in moderately and highly congested states, respectively. Results provide meaningful insights in terms of both independent and combined application of PC and efficient MP control. Under congested conditions, a properly selected subset of critical intersections with MP produces better performance than installing MP everywhere. Adding PC with MP creates even more significant improvements. Finally, detailed analysis of total remaining travel distance in multi-region networks is performed via microscopic simulation, in the scope of evaluating the benefits of utilizing the recently proposed M-Model, which is disconnected from the steady-state approximation of conventional PL model, in aggregated MFD-based network control applications. Results indicate significant potential improvement in terms of accuracy of prediction, especially in cases of highly-dynamic traffic evolution patterns.