Search-Based Model Optimization Using Model Transformations

Design-Space Exploration (DSE) and optimization look for a suitable and optimal candidate solution to a problem, with respect to a set of quality criteria, by searching through a space of possible solution designs. Search-Based Optimization (SBO) is a well-known technique for design-space exploration and optimization. Model-Driven Engineering (MDE) offers many benefits for creating a general approach to SBO, through a suitable problem representation. In MDE, model transformation is the preferred technique to manipulate models. The challenge thus lies in adapting model transformations to perform SBO tasks. In this paper, we demonstrate that multiple SBO techniques are easily incorporated into MDE. Through a non-trivial example of electrical circuit generation, we show how this approach can be applied, how it enables simple switching between different SBO approaches, and integrates domain knowledge, all within the modeling paradigm.