Short-Term Scientific Mission

From Models to the ROAR-NET API and Back

Main theme: Problem Modelling and User Experience
Grantee: Jens Kosiol, Philipps Universität Marburg, Marburg, Germany
Host: Steffen Zschaler, King's College London, London, United Kingdom

Start date: 2025-09-08
End date: 2025-09-19
Awarded: 2025-06-05
Report approved: 2025-10-31

Description

This STSM aims to contribute to the goal of creating a specification language for randomized optimization problems that allows for re-use across optimization approaches and tools. The immediate goal is to create a bidirectional translation between the ROAR-NET API and the model-driven approach to optimization (MDO), where domain-specific models (textual or visual) serve as specification mechanisms. A further goal is to derive, inspired by that translation, a translation of domain-specific models into more machine-oriented representations of solutions to tackle existing significant performance overheads of MDO. This work shall serve as a basis to, in the future, enhance existing MDO tooling in such a way that it serves as an implementation of the ROAR-NET API.

Achievements

During the STSM, a conceptual bridge between MDO and the ROAR-NET API has been established that shows that, in principle, the model-based approach to specifying optimization problems provides all operations necessary so that tools implementing the ROAR-NET API can be used as a solver. A publication on that bridge and a prototypical implementation of it are in planning. Furthermore, for better supporting the ROAR-NET API and as one ingredient of a future translation of models into more machine-oriented data structures, good progress was made on the automated construction of model transformation rules that preserve the feasibility of solutions (encoded as models).