Séminaire LARIS - Victor Pillac
A partir de 10h30 en salle 411 à l’ISTIA
Le 17 juin 2014
Victor Pillac : "A path-generation matheuristic for large scale evacuation planning".
Natural and man-made disasters, such as hurricanes, floods, bushfires, or industrial accidents, often affect large populated areas, threatening the lives and welfare of entire populations. In such events, a common contingency is to evacuate the persons at risk to different shelters and safe areas. Existing work in evacuation planning typically relies on free-flow models in which evacuees are dynamically routed in the network. In contrast, we propose an evacuation algorithm that follows recommended evacuation methodologies. It divides the evacuated area in evacuation zones, each being instructed to leave at a specific time and following a pre-defined route.
More specifically, it generates evacuation routes for each evacuation zone and uses a lexicographic objective function that first maximizes the number of evacuees reaching safety and then minimizes the total evacuation time. In contrast with the column generation technique, the proposed approach does not rely on the explicit pricing of new columns but instead exploits features of the incumbent solution to generate one or more columns in the master problem.
In this presentation, we will detail the algorithm, show how it can be used to plan the evacuation of 70,000 people in the Sydney area, and introduce a decision support system for evacuation planning.
Victor Pillac - chercheur au NICTA ((previously known as National ICT Australia Ltd) is Australia's Information and Communications Technology (ICT) Research Centre of Excellence) - a obtenu son doctorat sur les tournées de véhicules dynamique en 2012, sous la direction de Christelle Guéret, alors à l'Ecole des Mines de Nantes, et AndrésMedaglia, de la Universidad de Los Andes de Bogota. Ses projets de recherche se centrent autour de la gestion des catastrophes naturelles, et en particulier les problèmes d'évacuations.