A Cross-Layer Review Framework for Real-Time Decision-Making in UAV Swarm Systems: Evaluating Reliability, Autonomy, and Operational Efficiency

Unmanned aerial vehicle (UAV) systems substantially rely on efficient data processing. These mechanisms enable real-time collaboration and data-driven decision-making in dynamic environments. Although existing surveys address swarm architectures, learning methods, or communication strategies in isolation, they often lack an integrated perspective that links operational constraints, such as latency, energy consumption, scalability, and security, to appropriate data processing strategies. This paper presents a comprehensive review of data-processing architectures and swarm-learning methodologies for UAV swarm systems using a cross-layer analytical framework. The proposed framework supports practical decision-making by jointly evaluating reliability, autonomy, and operational efficiency in response to real-time mission needs. Based on an analysis of more than 85 studies, this study provides a structured comparison of centralized, decentralized, and hybrid architectures. Our review highlights significant developments in adaptive object detection, context-based situational awareness, blockchain-integrated Deep Learning (DL) frameworks, and decentralized control strategies. In addition, a mission-driven taxonomy of UAV-based applications was developed to emphasize the relationships between mission objectives, data-processing roles, and suitable learning strategies. It identifies the key challenges that hinder real-world deployment and presents a research roadmap for next-generation UAV systems that emphasizes adaptive communication, resource-aware distributed intelligence, and secure collaborative learning. Overall, this review provides system-level design guidelines for selecting appropriate data-processing and learning strategies, thereby enabling the development of scalable, resilient, and operationally efficient UAV swarm systems for practical deployment.