Combinatorial Optimization under Uncertainty (09w5097)

Objectives

Researchers have been interested in addressing the impact of uncertainty on optimal decision-making for several decades and even more. However, recent developments at the theoretical and algorithmic levels, as well as the advent of much more powerful computers, have now made it possible to tackle problems that are truly meaningful from a practical standpoint. This tendency is even stronger in the case of problems that are of a combinatorial nature, since these are much more difficult to solve.

A distinctive feature of combinatorial optimization under uncertainty is the existence of a several rather isolated communities, rooted in different disciplines (applied probabilities, computer science, engineering, operations research, etc.), pursuing seemingly unrelated research streams. While these communities often share application areas and problems, they seldom shares their main results or their techniques. The main objective of this workshop is to attempt to fill this gap by getting together some of the top researchers of the various communities involved in combinatorial optimization under uncertainty. Researchers involved in some problems that have received a lot of attention, such as stochastic vehicle routing, revenue management and call centers optimization will also be asked to participate in the workshop.

The proposed workshop is a follow up on a much smaller workshop (23 participants) that took place in Norway in 2006 and that dealt with ``Stochastics in logistics and transportation''. This workshop clearly showed the need for much more interaction and exchanges between the various communities who develop models and solution approaches for combinatorial problems under uncertainty. It also highlighted the relevance of going beyond the transportation and logistics application areas and to consider links between the various approaches at a more fundamental level. We believe that bridging the gap between the various communities alluded to earlier may have several consequences:

1) Development of deeper insights regarding the performance of the different approaches on various problem classes;

2) A sharper understanding of the most appropriate techniques in a specific context, thus leading to improved performances in practical settings;

3) Potential for putting together ``hybrid'' approaches that would exploit the strong points of several research streams.

As we mentioned earlier, in many critical problem settings, there is currently a very heavy trend towards accounting explicitly for the inherent uncertainty when one tries to optimize key decisions. There is thus a real need for a connected and more effective research communities in this area. We believe that the proposed workshop will address this pressing need.

As a final remark, it should be highlighted that the proposal for this workshop was met with a lot of enthusiasm from several of the possible participants listed (the 42 persons listed have all confirmed that they are interested in the proposed workshop and that they accept being listed as possible participants). Persons familiar with stochastic programming, dynamic programming, simulation-based optimization, robust optimization, or the application areas will have noticed that the workshop will indeed regroup a fair number of the top researchers in these areas.