The Algorithm Is Not the Answer: A Methodological Review of Metaheuristic Controller Tuning

Abstract

Metaheuristic optimization has become a widely adopted approach for controller parameter tuning in nonlinear, multi-objective engineering systems where classical methods fail. However, the rapid expansion of the literature has not been matched by methodological rigor, raising concerns about reproducibility and practical relevance. This review provides a critical synthesis of studies published between 2025 and 2026, organized thematically around controller architectures, optimizer families, and objective function design. Across five application domains, EV charging, power systems, robotics, process control, and marine systems, we identify persistent gaps: objective functions that disregard engineering priorities, weak constraint handling, unfair benchmarking, poor reproducibility, and insufficient validation. To move the field forward, we offer four prescriptions: benchmark relentlessly, validate physically, attribute honestly, and report completely. We conclude that meaningful progress requires shifting focus from algorithmic novelty to problem-structure-aware, methodologically credible optimization frameworks.