Unmanned aerial vehicles (UAVs) for smart agriculture with machine learning: A system-oriented review of methods, applications, and challenges

Rising food demand and climate variability are accelerating the adoption of smart agriculture (SA), where unmanned aerial vehicles (UAVs) coupled with machine learning (ML) provide on-demand, high-resolution information for agronomic decision-making. However, most existing reviews focus on UAV platforms, sensors, communication links, and ML algorithms in isolation, and rarely examine how these layers interact in operational agricultural settings. This study surveys ML-enabled UAV applications in agriculture published since 2020. It introduces a layered architecture that links UAV airframe and payload design, sensor configuration, navigation and positioning, and communication subsystems with an end-to-end ML pipeline for data acquisition, preprocessing, feature engineering, model training, validation, and deployment. The survey places particular emphasis on traditional ML methods that are well-suited to moderate-sized datasets, edge or near-edge deployment, and agronomic interpretability. Across major application areas, it critically assesses methodological robustness and transferability and clarifies how flight planning and sensing choices propagate to model reliability. Finally, the study identifies current challenges and provides actionable recommendations for integrating advanced UAV–ML technologies into future SA systems.