On integrating data uncertainty and multi-objective optimization: application to problems in disaster relief logistics

Description

Many optimization problems in the field of disaster relief logistics feature multiple objectives as well as parameter uncertainty. In this talk, we focus on the selection of distribution facilities, using data sets for slow-onset disasters such as droughts and sudden-onset disasters such as earthquakes. The considered concurrent objectives are cost and population coverage. Uncertain parameters are, e.g., the demand at the population centers or the capacities of the considered facilities. Several different approaches for dealing with data uncertainty exist. In this talk, we review two-stage stochastic programming, conditional value at risk, and scenario-based adjustable robust optimization. We establish a theoretical relationship between the latter two concepts and we combine the different approaches with criterion space search schemes, such as the balanced-box method or the epsilon-constraint scheme to account for the multi-objective nature of the problems. Finally, we also show how to integrate the L-shaped method into a bi-objective branch-and-bound framework to efficiently solve mixed integer bi-objective two-stage stochastic programs. We discuss the obtained results, both from a methodological as well as from a managerial perspective.

Period	25 Oct 2019
Event title	ÖGOR ATHEA Workshop
Event type	Conference
Location	AustriaShow on map