Multiple-objectives architecture optimization by composition of model transformations

Overview & Synopsis

In this thesis, we propose a new exploration approach to tackle design space exploration problems involving multiple conflicting non functional properties. More precisely, we propose the use of model transformation compositions to automate the production of architectural alternatives, and multiple-objective evolutionary algorithms to identify near-optimal architectural alternatives. Model transformations alternatives are mapped into evolutionary algorithms and combined with genetic operators such as mutation and crossover. Taking advantage of this contribution, we can (re)-use different model transformations, and thus solve different multiple-objective optimization problems. In addition to that, model transformations can be chained together in order to ease their maintainability and re-usability, and thus conceive more detailed and complex systems.