| Traditional models of traffic assignment usually assume that flow is assigned to shortest paths with respect to current (flow dependent) travel delays. In this thesis, we advocate the use of strategies for modeling traffic assignment. While this concept has been previously applied with success in the realm of urban transit assignment, we show that it can apply as well to priority networks involving arcs with rigid capacities.; In Part 1 of the thesis, we propose a static model for the passenger assignment where the sole congestion effects are induced by the link capacities and where priorities among users are explicitly taken into account. These priorities result from the fact that users that stay on board at a transfer node simply keep their seat, and thus have a natural priority over other commuters that try to board the bus at this node. In this model, users of the transportation network select, at their origin node, a travel strategy. A strategy associates with each node of the network an ordered set of successor nodes. At each node, users select the first available outgoing arc in this ordered set, where the probability that an arc be available is proportional to its capacity and inversely proportional to the number of users longing for it. The path used depends on the access probabilities, which are themselves functions of flows and capacities. The cost of a strategy is defined as the weighted cost of the path being used. At equilibrium, expected costs corresponding to any two used strategies associated with a given origin-destination pair should be equal.; In Part 2, we develop a new model for the dynamic traffic assignment where the behavior of the users is represented by the concept of strategies. We believe that this concept allows a better representation of user behavior in dynamic networks. The model provides a discrete-time description of the variations of flows on congested networks where the first-in first-out (FIFO) discipline is satisfied and where the capacity of the arcs is an integral part of the model. For any period of time and at each node, vehicles can either transshipped on travel arcs or wait until the next period in a waiting queue. This approximation of real-situations induces a synchronization of the strategic travel's users in the resulting space-time network. (Abstract shortened by UMI.)... |
