A survey on pioneering metaheuristic algorithms between 2019 and 2024
Tansel Dokeroglu, Deniz Canturk, Tayfun Kucukyilmaz
TL;DR
The paper surveys metaheuristic algorithms introduced from 2019 to 2024, screening over 150 methods and identifying 23 highly influential ones based on citations, problem diversity, accessibility, tunability, innovative mechanisms, and robustness to stagnation. It provides detailed analyses of these algorithms, their core ideas, and representative variants, and surveys state-of-the-art applications across engineering, ML, logistics, healthcare, and energy. The work highlights trends such as hybridization, machine-learning integration, and parallel implementations, while discussing open issues like algorithm similarity, parameter tuning, and binary encoding. The study aims to guide researchers and practitioners in selecting durable, effective methods for complex, real-world optimization tasks and to illuminate future research directions in this rapidly evolving field.
Abstract
This review examines over 150 new metaheuristics of the last six years (between 2019 and 2024), underscoring their profound influence and performance. Over the past three decades, more than 500 new metaheuristic algorithms have been proposed, with no slowdown in sight. An overwhelming abundance that complicates the process of selecting and assessing the most effective solutions for complex optimization challenges. Our evaluation centers on pivotal criteria, including annual citation metrics, the breadth of the addressed problem types, source code availability, user friendly parameter configurations, innovative mechanisms and operators, and approaches designed to mitigate traditional metaheuristic issues such as stagnation and premature convergence. We further explore recent high impact applications of the past six years' most influential 23 metahueristic algorithms, shedding light on their advantages and limitations, while identifying challenges and potential avenues for future research.