Resumen de A mesoscopic traffic simulation based dynamic traffic assignment
María Paz Linares Herreros 
In terms of sustainability, traffic is currently a significant challenge for urban areas, where the pollution, congestion and accidents are negative externalities which have strongly impacted the health and economy of cities. The increasing use of private vehicles has further exacerbated these problems. In this context, the development of new strategies and policies for sustainable urban transport has made transport planning more relevant than ever before. Mathematical models have helped greatly in identifying solutions, as well as in enriching the process of making decisions and planning. In particular, dynamic network models provide a means for representing dynamic traffic behavior; in other words, they provide a temporally coherent means for measuring the interactions between travel decisions, traffic flows, travel time and travel cost. This thesis focuses on dynamic traffic assignment (DTA) models. DTA has been studied extensively for decades, but much more so in the last twenty years since the emergence of Intelligent Transport Systems (ITS). The objective of this research is to study and analyze the prospects for improving this problem. In an operational context, the objective of DTA models is to represent the evolution of traffic on a road network as conditions change. They seek to describe the assignment of the demand on different paths which connect every OD pair in a state of equilibrium. The behaviour following each individual decision during a trip is a time-dependent generalization of Wardrop's First Principle, the Dynamic User Equilibrium (DUE). This hypothesis is based on the following idea: When current travel times are equal and minimal for vehicles that depart within the same time interval , the dynamic traffic flow through the network is in a DUE state based on travel times for each OD pair at each instant of time ([ran-1996]). This work begins with the time-continuous variational inequalities model proposed by [friesz-1993] for solving the DUE problem. Different solutions can be used on the proposed DUE formulation. On the one hand, there are the so-called analytical approaches which use known mathematical optimization techniques for solving the problem directly. On the other hand, there are simulation-based formulations that approximate heuristic solutions at a reasonable computational cost. While analytical models concentrate mainly on deriving theoretical insights, simulation-based models focus on trying to build practical models for deployment in real networks. Thus, because the simulation-based formulation holds the most promise, we work on that approach in this thesis. In the field of simulation-based DTA models, significant progress has been made by many researchers in recent decades. Our simulation-based formulation separates the proposed iterative process into two main components: - A method for determining the new time dependent path flows by using the travel times on these paths experienced in the previous iteration. - A dynamic network loading (DNL) method, which determines how these paths flow propagate along the corresponding paths. However, it is important to note that not all computer implementations based on this algorithmic framework provide solutions that obtain DUE. Therefore, while we analyze both proposals in this thesis we focus on the preventive methods of flow reassignment because only those can guarantee DUE solutions. Our proposed simulation-based DTA method requires a DNL component that can reproduce different vehicle classes, traffic light controls and lane changes. Therefore, this thesis develops a new multilane multiclass mesoscopic simulation model with these characteristics, which is embedded into the proposed DUE framework. Finally, the developed mesoscopic simulation-based DTA approach is validated accordingly. The results obtained from the computational experiments demonstrate that the developed methods perform well.
