Présentation sur YouTube.

In this talk, we study a congested facility location problem in which each facility is modeled as an M/M/c queue. We assume that users are self-interested and choose which facility to patronize so as to minimize their expected total time, defined as the sum of travel time and waiting time. This behavior leads to a bilevel problem in which the upper-level decision maker, the network designer, seeks an effective configuration of facilities, while the lower-level problem corresponds to a user equilibrium model. We propose a flexible single-level reformulation that can accommodate various objectives related to overall system efficiency as well as fairness. The resulting reformulation is a convex mixed-integer nonlinear program that can be solved efficiently.