K-center problems are a kind of prototype for facility location problems. They are very simple in their formulation, but (NP-) hard to be solved, even when fixing the dimension or k. We present a major (theoretical and practical) break-through result, a draw back, and why we think the k-center problem is still worth to be further investigated.
Alexandre Forel: "Dynamic stochastic capacitated lot-sizing problem: A scenario-based pre-emptive approach to mitigate nervousness"
In make-to-stock environments, production planners have to determine orders and production quantities using uncertain forecasts. Lot-sizing decision are made to minimize production costs while ensuring customer satisfaction. The process is usually modelled as a MILP and implemented in a rolling horizon fashion. An important problem that managers face when determining production plans lies in the nervousness induced by frequent changes in consecutive plans. Popular methods used to mitigate nervousness penalize or prohibit deviations from previously calculated plans: these methods are reactive. On the contrary, we aim at finding a solution method that reduces planning instability by explicitly considering that the plan may be revised in future periods. Using a scenario-based representation of the uncertainty, we investigate a pre-emptive method to mitigate nervousness and find compromises between flexibility and stability. A piecewise-linear approximation of the stochastic problem is also implemented as a benchmark for the optimal stochastic solution without pre-emptive nervousness considerations. Preliminary analysis of the impact of uncertainty and errors of the demand forecasts are conducted to assess the applicability of the methods in real-world settings.
Venue: TUM Campus Garching Room 01.10.011
Date: Monday, June 11th, 2018, 14:00