Lightweight AI for Drones: A Survey

This paper explores the transformative potential of Lightweight AI techniques in enhancing the capabilities of Unmanned Aerial Vehicles (UAVs). As the demand for real-time processing and efficient data management grows, traditional AI models face significant limitations due to their high computational and resource requirements. We discuss key Lightweight AI techniques, including Tiny Machine Learning (TinyML), model compression methods, edge computing, transfer learning, and federated learning, which collectively enable more efficient UAV operations. The paper further examines various applications of Lightweight AI in drones, highlighting its role in pandemic response, object detection, vegetation segmentation, forest-fire detection, real-time path planning, and autonomous landing. Each application demonstrates how lightweight models can improve decision-making and operational efficiency in dynamic environments. We also address the challenges faced in implementing these technologies, such as resource constraints and the need for robust real-time processing. Finally, we outline future directions for research and development in Lightweight AI, emphasizing the necessity for continued innovation to fully leverage UAV potential across diverse sectors. This comprehensive analysis underscores the critical role of Lightweight AI in advancing UAV technology and enhancing their functionalities in various practical applications.