Every second counts for search and rescue: A systematic review of TinyML drone

This study explores the transformative potential of TinyML in unmanned aerial vehicles (UAVs) to address key inefficiencies in traditional search and rescue (SAR) operations, especially in the context of increasingly severe climate-related disasters. By analyzing peer-reviewed studies in major technical databases via the PRISMA guidelines, this work highlights advancements in edge computing, swarm intelligence, and multisensory integration, with a focus on fundamental contributions in embedded AI and autonomous navigation. UAVs supported by TinyML can achieve low-latency and energy-efficient real-time processing, thereby enhancing the efficiency of disaster relief operations in harsh environments. This study emphasizes the need to create synthetic datasets for underrepresented scenarios, conduct robustness tests under extreme conditions, and adopt privacy-focused decentralized learning. It connects technological progress with ethical issues such as monitoring risks and equitable access to disaster technologies. Future research directions can overcome current limitations, including insufficient validation in practical applications, fragmented policies, and high costs in resource-poor regions, through interdisciplinary collaboration, transforming theoretical advancements into scalable and socially responsible TinyML-UAV system solutions.