Bi-objective facility location in the presence of uncertainty: An evaluation of stochastic and robust modeling approaches

Description

Multiple and often conflicting criteria need to be taken into account in real world problems. Moreover, due to dealing with data in a non-precise real world, considering uncertainty is of vital importance. To cope with uncertainty in optimization problems, many different approaches have been presented in the literature. The most widely used ones are stochastic optimization including concepts such as the expected value, chance constraints or risk measure, and robust optimization, including prominent concepts such as minmax robustness or adaptive robust optimization. This paper aims at investigating bi-objective modeling frameworks for an uncertain location-allocation model to design a humanitarian aid delivery network in disastrous situations. In order to find an efficient and reliable methodology to solve the problem, we consider slow-onset as well as sudden-onset disaster settings which differ in the sources of uncertainty. We use three different approaches to model uncertainty: scenario-based two-stage stochastic optimization, minmax robust optimization and adaptive robust optimization. To deal with the bi-objective nature of the problem, all three approaches are embedded into criterion space search methods, namely the well-known e-constraint method and the recently introduced balanced box method. We evaluate and compare the performance of the applied approaches on data sets derived from real world case studies.

Period	05 Sept 2019
Event title	Operation Research 2019
Event type	Conference
Location	GermanyShow on map