腾讯会议号：710 260 084
报告人简介：
Francisco Saldanhada Gamais professor of Operations Research in the Department of Statistics and Operations Research at the Faculty of Science, University of Lisbon, Portugal, where here rceived his PhD in 2002. He has extensively published papers in scientific international journals mostly in the areas of location analysis, supply chain management, logistics and combinatorial optimization. Recently, together with GilbertLaporte and Stefan Nickelco-edited the second edition of the volume Location Sciencepublished by Springer International Publishing (26chapters, 767pages). Awarded several prizesand honors. Among those, the EURO prize for the best EJOR review paper (2012) and the Else vierprize for the EJOR top cited article 2007–2011 (2012), both with the paper entitled Facility location and supply chain management—A reviewwritten together with Stefan Nickel and Teresa Melo. Member of innumerable scientific committees of international conferences and others cientific events. Member of various international scientific organizations such as INFORMS—Institute for the Operations Research and Management Science, USA, CMAFcIO—Centro de Matemática Aplicações Fundamentais e Investigação Operacional da Fundação da Faculdade de Ciências, University of Lisbon, ECCO—European Chapteron Combinatorial Optimization, EWGSO—Working Group on Stochastic Optimization, SOLA—INFORMS Sectionon Location Analysis, and EWGLA—EURO Working Groupon Locational Analysis, of which he is one of the past coordinators. Currently he is Editor-in-Chief of Computers & Operations Researc has well as member of the Editorial Advisory Boards of the Journal of the Operational Research Society (UK) and Operations Research Perspectives. His research interests includes to chastic mixed-integer optimization, location theory and project scheduling.
报告摘要：
Districting Problems (DPs) aim at partitioning a set of basic geographic areas, called Territorial Units (TUs), into a set of largerclusters, calleddistricts. This is done according to some planning criteria that typically refer to balancing, contiguity and compactness (KalcsicsandRios-Mercado, 2019). These are problems have a wide range of applications that include political districting, strategic service planning and management (e.g. inhealthcare), school systems, energy and power distribution networks, design of police districts, waste collection, transportation, design of commercial are as to assign sales forces, and distribution logistics.
One aspect of practical relevance in DPs concerns the need to cope with changing demand. Depending on the specific problem considered, different possibilities emerge. One is to assume are active posture and solve a so-called redistricting problem (Brunoetal., 2016). Another is to embed time in the optimization models. This is a possibility if accurate forecasts for the demand are available (Bender, 2018). Finally, if demand changes are uncertain, then embedding such uncertainty in a mathematical model may be desirable. Assuming that uncertainty can be measured using some known probability function, a stochastic programming model emerges asappropriate (CorreiaandSaldanha-da-Gama, 2019).
In this presentation, a family of stochastic districting problems is discussed. Demand is assumed to be represented by a random vector with a given joint cumulative distribution function. A two-stage mixed-integer stochastic programming model is proposed. The first stage comprises the decision about the initial territory design, i.e., the districts are defined and all the TUs assigned to one and exactly one of them. In the second stage---after demand becomes known---balancing requirements are to be met. This is ensured by means of two recourse actions: out sourcing and reassignment of TUs. The objective function accounts for the total expected cost that includes the cost for the first-stage territory design plus the expected cost incurred by outsourcing and reassignment in the second stage. The (re)assignment costs are associated with the distances between territory units, which is a means to put emphasis on the compactness of the solution. The initial model is extended in different ways to account for aspects of practical relevance such as a maximum desirable dispersion, reallocation constraints, or similarity of the second-stage solution w.r.t. the first stage one. The new modeling framework proposed was tested computationally using instances built using real geographical data. The obtained results are reported.