A comparator-based constraint handling technique for evolutionary algorithms

Hikaru Takami, Shigeru Obayashi

Research output: Contribution to journalArticlepeer-review

1 Citation (Scopus)

Abstract

Constraint handling is a key task for the successful optimization of design parameters in industrial design problems. This paper proposes a comparator-based constraint handling technique, called the More Less-Violations Method (MLVM), for solving real constrained optimization problems using evolutionary algorithms. The structure of the MLVM is simple and it can easily be integrated into conventional evolutionary algorithms. In the proposed method, constraint weights represent the level of importance of each constraint, enabling evolutionary compliance prioritization. Moreover, an acceptable region formed by the constraint tolerances allows trade-offs between objectives and constraints while preserving diverse solutions and improving optimization performance. These elements enable the appropriate design of industrial optimization problems. An application of this method to problems without constraint tolerances is also proposed. The JAXA/Mazda benchmark problem, developed on a real-world constrained design optimization dataset, is used to assess the performance of the MLVM. The results indicate that the MLVM realizes encouraging optimization performance.

Original languageEnglish
Article number055229
JournalAIP Advances
Volume12
Issue number5
DOIs
Publication statusPublished - 2022 May 1

Fingerprint

Dive into the research topics of 'A comparator-based constraint handling technique for evolutionary algorithms'. Together they form a unique fingerprint.

Cite this