Characterization of constrained continuous multiobjective optimization problems
A. Vodopija (Supervised by B. Filipič)
Within the doctoral research project, we aim at characterizing constrained continuous multiobjective optimization problems (CMOPs) to facilitate the preparation of an experimental setup for benchmarking constrained multiobjective optimization algorithms. We address this task by generalizing the landscape analysis to constrained multiobjective optimization and designing novel landscape features for CMOPs. Furthermore, we propose a novel algorithm performance approach specialized for constrained multiobjective optimization. Finally, using the developed techniques, we comprehensively analyze the current test suites of CMOPs and provide novel insight for benchmark and algorithm designers.