Variable-Sized Robustness at Aston

Last week I returned to Birmingham’s Aston University (having been there in March to talk about optimisation methods in evacuation planning) and joined the first-of-its-kind IMA and OR Society Conference on Mathematics of Operational Research. That’s kind of awkward name for a great idea. The IMA, that is the Institute of mathematics and its applications, a society for the promotion of mathematical culture, and the OR Society have plenty of things in common, and organising a joint conference seems quite natural in hindsight. I’ve been to few other conferences where I enjoyed every single talk.

I used the opportunity to talk about variable-sized robustness, which has become an area of great interest for me recently. The basic idea, as we explored it over two papers (1,2), is that uncertainty sets in robust optimisation are typically not known to the decision maker. Usually only raw data is available, and a suitable uncertainty set needs to be determined before any robust model can actually start its work. In variable-sized robustness, we assume that only the desired shape of the set is known, but not its size. This leads to some new and challenging optimisation problems with close ties to multi-objecitve optimisation.

The IMA/OR conference was well attended and we can only hope to see more events of the same kind in the future!