Work

Feasible Regions Identification based on Historical Solutions for Constrained Optimization Problems

Year: 2025

Type: conference-paper

Abstract: The presence of constraints often leads to the formation of narrow and fragmented feasible regions within the search region, presenting significant challenges for optimization problem-solving. This paper introduces a novel approach, Feasible Regions Identification based on Historical Solutions (FRIHS), designed to address these challenges. FRIHS leverages previously evaluated solutions to partition the search region into ε-feasible and ε-infeasible regions. Additionally, by analyzing the correlations among constraints, they are reformulated as auxiliary objectives, effectively transforming the constrained optimization problem into a constrained multi-objective optimization problem. The method employs the classical evolutionary algorithm Differential Evolution and the multi-objective method NSGA-III to search the most promising feasible regions. The effectiveness of FRIHS is evaluated through a comparative analysis with five advanced constraint-handling algorithms across a benchmark test suite. Experimental results indicate that the proposed approach demonstrates competitive performance on the test problems.

Source: IEEE Congress on Evolutionary Computation (CEC)

DOI: 10.1109/cec65147.2025.11043100

Authors: Mengli Shan, Changhe Li, Mai Peng, Xiaobo Liu, Michalis Mavrovouniotis, Shengxiang Yang

Institutions: China University of Geosciences, ERATOSTHENES Centre of Excellence, De Montfort University

Language: English

Topic: Metaheuristic Optimization Algorithms Research

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: green