Metaheuristic algorithms for complex optimization: a critical review of foundations, hybrid strategies, and surrogate modeling

Abstract

Metaheuristic algorithms have emerged as indispensable tools for solving complex optimization problems across engineering, logistics, and data-driven decision-making domains. Unlike traditional methods that require strict assumptions or derivative information, metaheuristics provide flexible frameworks capable of delivering high-quality, near-optimal solutions in high-dimensional, multimodal, and constrained search spaces. This paper presents a comprehensive and critical survey of metaheuristic algorithms, offering a structured taxonomy that spans single-solution and population-based methods as well as classical and emerging approaches. Special attention is devoted to hybridization strategies and surrogate-assisted techniques, which are increasingly employed to enhance convergence speed, robustness, and scalability in computationally expensive environments. Drawing upon recent studies, we analyze strengths, limitations, and parameterization challenges while identifying persistent gaps in reproducibility, adaptive control, and large-scale applicability. Finally, we propose a forward-looking roadmap emphasizing adaptive parameter tuning, ensemble surrogates, and parallel or distributed computing to guide the next generation of metaheuristic optimization frameworks. This synthesis provides researchers and practitioners with an authoritative reference for selecting, adapting, and innovating metaheuristic algorithms in diverse application contexts.