A Stigmergy-Based Island Model for Dynamic Evaluation of Constraint-Handling Techniques and Differential Evolution Algorithms

A nontrivial task is to choose an evolutionary algorithm (EA) and a constraint-handling technique, among countless possibilities, leading to a robust and efficient mechanism to solve constrained optimization problems (COPs). The main motivation of this article is to automate the decision on the EA coupled with a constraint-handling technique able to find the best performance in solving COPs. From this issue, this work proposes a methodology based on the stigmergy island model (IM), where EAs and constraint-handling techniques are distributed throughout the islands, and their performances are dynamically evaluated along the evolutionary process to guide the migration of solutions between them. The stigmergy IM is a strategy for implementing the hybrid IMs dynamically and adaptively, inspired by the natural phenomenon of stigmergy. According to the EAs applied on the islands, the weighted connections and distribution of solutions between islands are adjusted over migrations. The proposed model, which is the main contribution of this work, identifies and explores the most suitable algorithm for COPs among those applied in the initial topology. In this sense, different hybrid alternatives based on different differential evolution algorithms and constraint-handling techniques are presented to evaluate the performance of the proposed IM.