What Evolutionary Algorithm(s) Should We Use? Towards Preliminary Guidelines from an Exploratory Analysis of Foraging Problems

Researchers often struggle to identify a suitable method for addressing problems. This is particularly evident when there are too many options but no general criteria to guide the choice. In the context of Evolutionary Algorithms (EAs), the literature provides a wide range of viable alternatives, but few works have focused on defining guidelines to choose one algorithm over another. As a result, a common approach is to rely uncritically on existing studies, although the risk of failures and/or sub-optimal performance cannot be excluded. This research helps address this gap by providing a list of practical guidelines for selecting the most appropriate EA. Specifically, we perform a preliminary investigation on three Multi-Agent System (MAS) foraging problems, comparing six EAs with different levels of complexity. Moreover, we identify task-specific metrics to characterize and discriminate the problems under consideration. Our analysis pinpoints the EA(s) that yield the best performance for each problem and reveals relationships between the proposed metrics and algorithmic effectiveness, thereby providing guidelines for the selection of suitable algorithms.