REAL-TIME PATH PLANNING FOR UAV WITH A FOCUS ON SAFETY

The increasing number of UAVs raises concerns about the safety of airspace and the people near operational sites. In Brazil, the Department of Airspace Control (DECEA) and the National Civil Aviation Agency (ANAC) have established a minimum distance of 30 meters between the aircraft and people, buildings, properties, and animals. To ensure safe UAVs routes and compliance with flight conditions, this paper presents an online path planning method that realizes for real-time route replanning if people are detected along the path. An offline route is predefined to avoid static obstacles mapped from high-resolution Remote Sensing products. Field experiments with the DJI SDK demonstrate that the offline route successfully avoids static obstacles with precision. Upon detecting a person with the YOLOv10s model (average precision of 65.4%) and determining their coordinates using Monoplotting, the route is replanned in 7.78 ms, quickly avoiding dynamic obstacles – thus enabling real-time processing for UAV field operations.