EURO 2025 Leeds
Abstract Submission

2201. Two-stage robust multiobjective optimization with an interactive method

Invited abstract in session TC-51: Multiobjective Decision Making, stream Multiobjective and vector optimization.

Tuesday, 12:30-14:00
Room: Parkinson B22

Authors (first author is the speaker)

1. Juho Roponen
Faculty of Information Technology, University of Jyvaskyla
2. Babooshka Shavazipour
University of Jyvaskyla
3. Margaret Wiecek
School of Mathematical and Statistical Sciences, Clemson University

Abstract

Most areas of human activity involve trade-offs and uncertainty. Trade-offs make optimization difficult, because in the presence of multiple conflicting objectives, there is no individual optimal solution, but instead multiple – possibly infinitely many – Pareto optimal solutions. Uncertainty also makes optimization more difficult, because often no individual solution is going to be clearly the best for every possible use case or future scenario. When multiple objectives and uncertainty are considered simultaneously, the relative importance of different objectives may change based on the achievable ranges of the objectives in different scenarios. We present an interactive approach to solving two-stage robust multiobjective optimization problems. The advantage of solving the optimization problem interactively is that the decision maker has an opportunity to learn about trade-offs and uncertainties involved and redefine their preferences during the solution process. The updated preferences are expressed in a scalarized min-max-min problem which is reformulated into a minimization problem. By iteratively updating preferences and solving the scalarized problem, the decision maker can find a preferred solution. We apply this solution approach to a forestry problem with initial management decisions, after which, one of finitely many potential climate scenarios is realized, and a second round of management decisions follows.

Keywords

Status: accepted

Back to the list of papers