Original scientific paper
https://doi.org/10.20532/cit.2026.1006008
A Dynamic Multi-objective Optimization Algorithm for Black-Box Optimization in Dynamic Irregular Environment
Sanyi Li
; School of Electrical and Information Engineering, Zhengzhou University of Light Industry, Zhengzhou, China
*
Shuang Liu
; College of Mechanical and Electrical Engineering, Linzhou College of Architectural Technology, Anyang Henan, China
Wenjie Hou
; School of Electrical and Information Engineering, Zhengzhou University of Light Industry, Zhengzhou, China
Yuanyuan Zhang
; College of Mechanical and Marine Engineering, Beibu Gulf University, Qinzhou Guangxi, China
* Corresponding author.
Abstract
This paper proposes a novel initial population prediction algorithm (IPPA) for solving irregular black-box dynamic multi-objective optimization problems. In the phase of environmental change detection, IPPA find representative solutions based on a clustering algorithm and form environment vector to judge whether the environment has changed. In the population initialization stage, the initial population is generated by three mechanisms. First, a feedforward neural network (FNN) is trained with historical population information, and part of the initial population is generated by FNN. Second, when the new environment is similar to the last environment, part of the elite population in the last environment is retained. Finally, some populations are randomly generated to maintain population diversity. Since IPPA does not require the use of explicit objective functions and recent solutions, it can solve black-box dynamic multi-objective optimization problems with drastic and irregularly changing Pareto set. The proposed algorithm is compared with other state-of-the-art algorithms on a number of benchmark problems. Experimental results show that the proposed IPPA algorithm is effective for irregular blackbox dynamic multi-objective optimization problems.
Keywords
dynamic multi-objective optimization; pareto Set, neural network; prediction; irregular environment
Hrčak ID:
346533
URI
Publication date:
7.4.2026.
Visits: 32 *