An introduction to robust optimisation
Thursday 18 March 2010, 14:00
LT5, Management School
University of Bologna
Abstract: For many years, it was assumed that the first thing to consider when tackling an optimisation problem is the complete knowledge of the data. A determinist model was not regarded as a matter of choice, but as the basis of all possible treatments. Yet we should think twice before making such an assumption, since there are many examples that show the major hazards of a deterministic model. Real-world optimization problems are often subject to uncertainty: measurement errors, implementation errors, system disturbances, or incomplete available data at the moment of planning. All of these factors can make the nominal solution completely meaningless from a practical point of view, and there is hardly any real context which is immune from these dangers. For this reason, many approaches to deal with uncertainty have been developed. Among them, Robust Optimisation has received a consistent share of attention from the research community over the last ten years. This talk will introduce the non-expert to Robust Optimisation, taking three steps: firstly, it will overview the main non-robust approaches to deal with uncertainty; secondly, starting from the work of Soyster (1972), it will show the development of the field of Robust Optimisation; finally, it will point out the limitations of ‘classical’ robustness and the more recent extensions towards robust multi-stage optimisation.
About the speaker: After doing a masters in computer science engineering and gaining professional accreditation as an engineer, Laura Galli gained a PhD in Control System Engineering and Operational Research at the University of Bologna, Italy. Her thesis was concerned with applications of robust optimisation in the railway sector. Dr Galli is currently working at the University of Bologna as a post-doctoral research associate.