2918. Beyond monolithic solvers: Solving multidisciplinary optimization problems in a cooperative fashion
Invited abstract in session TB-51: Multiobjective discrete optimization, stream Multiobjective and vector optimization.
Tuesday, 10:30-12:00Room: Parkinson B22
Authors (first author is the speaker)
| 1. | Dimitri Rusin
|
| Alliance Manchester Business School, University of Manchester | |
| 2. | Manuel López-Ibáñez
|
| University of Manchester | |
| 3. | Joshua Knowles
|
| Invenia Labs | |
| 4. | Ali Hassanzadeh
|
| Alliance Manchester Business School, University of Manchester |
Abstract
High-stakes applications of multiobjective discrete optimization often require that a system be fully optimized without knowing its full formulation. Instead, the formulation is fragmented into two or more components that depend on each other, yet must be optimized consecutively or in parallel.
Such applications can be formulated as multidisciplinary optimization models, also known as interwoven systems, which can be naturally decomposed into (at least) two components. We seek a set of solutions that is Pareto optimal with respect to both components.
Two components interact: On the one hand, the first component can propose a baseline solution that will allow the second component to be optimized independently. On the other hand, if the first component slightly tweaks its baseline solution, the second component's optimized solution can become suboptimal.
In this talk, we study this interaction in the context of the Travelling Thief Problem. The Travelling Thief Problem famously models an interaction between the Travelling Salesman Problem and the Knapsack Problem. Even if we can solve both of these individual single-objective problems efficiently, it is not obvious how to integrate these efficient solvers into an efficient solver of the aggregated multiobjective optimization problem.
We present first results of our study of the relationship between the number of interaction rounds and the achieved Pareto efficiency gap.
Keywords
- Combinatorial Optimization
- Multi-Objective Decision Making
- Game Theory
Status: accepted
Back to the list of papers