Robust optimization is becoming more and more popular in Operations Research as a tool to cope with uncertainty that is present in many modeling situations, but generally neglected. Actually there are many major obstacles to a proper treatment of uncertainty. For instance, the information on the probability distributions is often partial and incomplete, and even if the distributions are known, computing probabilities and expectations turns out to be a numerical challenge. Robust optimization acknowledges these facts and proposes an approach which enables efficient and numerically tractable computation of ``reasonable' ' solutions on the basis of limited information. In this presentation, we shall first review a few basic concepts and show how they often lead to relatively simple formulations. We shall illustrate the talk with two examples. The first one concerns a long-term energy planning model of the Markal family and aims at increasing European energy security when supplies are uncertain. The second one proposes a way to handle uncertain demand forcasts in the capacity expansion problem for telecommunication networks.
| Subject : | : | typdoc_1455 |
| Language of text | : | French |
| Topics | : | Stochastic and robust optimization, online optimization, queuing theory and simulation, machine learning and statistical methods |
| PDF version | : |  PDF version |
