Dynamic traffic assignment models for disrupted networks

Transportation infrastructure systems are one of the cornerstones on which modern societies are founded. They allow the movement of people and goods by enabling business activities, the setting up of supply chains, and they provide access to vital resources and services. It is commonly believed that due to their vast scale and complexity, transportation systems are among the most vulnerable infrastructures in the occurrence of a disruption, i.e. an event that involves extensive damage to people or physical facilities. The growing awareness about this issue in recent years has led to a growing body of literature on the topic of performance evaluation of transportation networks when affected by disruptive events, aimed at providing adequate estimations of network operability in such contexts. A peculiarity of transportation is to be a socio-technical system where the transportation supply, represented by the infrastructure and related services, interacts with the transportation demand, consisting of all those individuals who access the infrastructure at any given time. This property makes such systems inherently complex and consequently an analysis of their vulnerability in the face of disruptive events needs to be able to account for these interactions. The aim of the present thesis is therefore to develop methodologies capable of convincingly portraying the reaction of such systems in the face of disruptive events by modeling the dynamics that emerges between users and the infrastructure. It is reasonable to assume that travelers due to changed system conditions will adapt their behavior to some degree in order to mitigate the consequences of such events. In this regard, three modeling approaches are presented in this manuscript to address the need to represent this reaction phenomenon. The first approach involves the use of an inter-period traffic assignment model able to represent the evolution of users' mobility choices in a dynamic context. For each period, the users' reaction is estimated by solving an assignment model thus computing the optimal flow distribution given the current congestion conditions. User habits are taken into account by appropriately limiting the extent of flow redistributions in order to represent the gradual adaptation of the system to the new situation. Large perturbations can trigger modal shift phenomena between one transport sub-system and another. In this regard, a multi-modal multi-class scenario analysis model is then presented. Railway and road transport sub-networks are thus embedded into an extended hyper-network to model flow exchanges between this two sub-systems. Class-specific assignment models are employed to determine the choice behavior for passenger flows and freight flows. The results of these choices are then routed through the network by means of a discrete-time dynamic flow model. Finally, the idea that users' behavior may be influenced by their habits is further explored within a path-based inter-period assignment model. It is suggested that users' route choice process is not only influenced by the travel costs of available alternatives but also by users' familiarity with them. More specifically, if changing traffic conditions suddenly make a specific route disadvantageous, users will tend to prefer those that are most topologically similar to the one they are abandoning. This assumption is then investigated by demonstrating that it implies considering a rationally bounded user choice process. The steady state reached by the system as a result of the equilibration process is then detailed and a rigorous proof is provided to show that it is equivalent to a Boundedly Rational User Equilibrium. All three approaches has been successfully applied on appropriate test networks where a disruption is simulated by altering the network topology. These models can provide an important contribution to transportation network vulnerability and resilience analyses willing to take into account the interaction between the infrastructure and users.