A Comprehensive Review of Computational Intelligence in Modern Electronic Warfare: From Deterministic Approaches to Hybrid Meta-Heuristic Optimization

Abstract

Electronic warfare (EW) systems are a critical component of the modern battlefield, enabling strategic management of the electromagnetic spectrum and establishing operational superiority. Although traditional deterministic methods provide foundational mathematical frameworks for radar detection and signal analysis, they remain limited in flexibility and adaptability when faced with dynamic and cognitive threat models. This study systematically and comparatively examines signal processing, optimization, and learning-based techniques employed in modern EW systems, drawing upon 38 original research works identified in the literature. It categorizes them under four principal methodological classes: traditional methods, machine learning, deep learning, and metaheuristic optimization. The analyses reveal that deep learning models achieve superior performance, with classification accuracies reaching 98.08%, particularly under high signal-to-noise ratio (SNR) conditions. At the same time, metaheuristic algorithms yield improvements of up to 18.27% over conventional approaches in non-convex, complex optimization problems such as focused energy delivery (FED) and waveform design. Nevertheless, noise sensitivity under low-SNR conditions, the inability to guarantee global optimality, and real-time implementation constraints are identified as principal open research challenges in light of the reviewed studies. The findings underscore the necessity of hybrid approaches that reconcile deterministic reliability with adaptive learning capacity, as well as the integration of explainable artificial intelligence, to enhance the operational dependability of modern EW systems.