Solving large-scale, mid-term planning problems under multiple objectives – A contribution to VeRoLog 2017 optimization competition

VCLA and DBAI hosted a talk by Martin Josef Geige

ABSTRACT

The talk presents a solution approach for the recent implementation  challenge of the EURO Working Ground on Vehicle Routing and Logistics Optimization, a multi-objective, multi-period vehicle routing problem with pickups, deliveries, and possible transshipments. Important objectives comprise the classical minimization of the length of the routes, but also other considerations such as the fleet size, as well as the product levels over a longer planning horizon. Between the objectives, considerable trade-offs exists that result in a true multi-criteria view on the problem at hand (although the objectives are, for the purpose of the challenge, combined in an overall cost function).

Reflecting the different scenarios and data sets, a general solution framework has been developed that first allows the setting of the relative importance of criteria. Then, the data set is solved by means of several techniques: (i) Variable Neighborhood Search is used for the routing of the vehicles; (ii) lower bounds have been found and computed for the product levels: they guide the search towards qualitatively good solutions. On a lower level, there are several interesting subproblems that have been tackled, such as quick feasibility checks and problem-specific data structures that generalize the combination of pickups and deliveries.

Biography:

Martin Josef Geiger is a professor of Business Administration/Logistics Management at the Helmut Schmidt University/University of the Federal Armed Forces Hamburg, Germany. He holds a Doctorate degree and a post-doctoral Habilitation/ Venia Legendi in Business Administration, both from the University of Hohenheim, Stuttgart, Germany. His research interests primarily lie in quantitative planning problems with applications in logistics, routing, and scheduling. In this context, he works on understanding and applying modern (meta-)heuristics to such problems.