An Adaptive Dynamic Migration-based Constrained Multiobjective State Transition Algorithm

Complex constrained multi-objective optimization problems are often characterized by narrow and disconnected feasible regions, making it challenging to obtain the optimal constrained Pareto front. To improve feasibility, diversity, and convergence in the optimization process, this paper introduces a...

Full description

Saved in:
Bibliographic Details
Published in:Youth Academic Annual Conference of Chinese Association of Automation (Online) pp. 962 - 968
Main Authors: Chen, Xiaolong, Han, Jie, Zhu, Liang, Yang, Chunhua, Wang, Xiaoli
Format: Conference Proceeding
Language:English
Published: IEEE 07.06.2024
Subjects:
adaptive dynamic migration
Automation
Benchmark testing
Constrained multiobjective optimization
Constraint handling
constraint handling techniques
Heuristic algorithms
Optimization
Sociology
state transition algorithm
Statistics
ISSN:2837-8601
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Complex constrained multi-objective optimization problems are often characterized by narrow and disconnected feasible regions, making it challenging to obtain the optimal constrained Pareto front. To improve feasibility, diversity, and convergence in the optimization process, this paper introduces an adaptive dynamic migration-based constrained multi-objective state transition algorithm. The approach starts with an individual-based state transition algorithm, which searches for candidate solutions across the entire search space. Subsequently, these candidate solutions are evaluated using an adaptive dynamic migration constraint handling technique, which enhances the convergence of candidate solutions in the early stages of optimization and ensures their feasibility in the later stages. The migration process is adaptively adjusted based on the diversity and convergence of the population to maintain a balance among feasibility, diversity, and convergence. Experiments conducted on 14 benchmark problems reveal that the proposed method demonstrates superior or, at the very least, competitive performance when compared to other well-established methods.
ISSN:2837-8601
DOI:10.1109/YAC63405.2024.10598512