A dynamic multi-objective optimization evolutionary algorithm for complex environmental changes

Abstract

Dynamic multi-objective optimization problems (DMOPs) have attracted more and more research in the field of evolutionary computation community in recent years. Unlike most existing approaches just for solving a single change type, we propose a novel dynamic diversity introduction strategy (DDIS), which aims to solve DMOPs with mixed complex environmental changes. Two types of change intensity are presented to jointly determine the proportion of diversity introduction and whether the change type is drastic or slow, and then the inverse modeling and partial population random initialization are served as diversity introduction strategies to respond to environmental changes respectively. The proposed DDIS is incorporated into the multi-objective evolutionary algorithm based on decomposition (MOEA/D) framework, called DDIS-MOEA/D. For verifying the performance of DDIS, three different mixed change types are constructed by varying severity or frequency of changes and then the proposed algorithm is tested on GTA benchmark problems under the three dynamic characteristics. Experimental results confirm that the proposed approach can successfully identify different change types and dynamically track and adapt complex environmental changes.