Development of a drone-based rescue platform with intelligent human detection and multi-payload delivery mechanism

Floods often create emergencies in which access to victims is severely constrained by fast-moving water, deep terrain, and hazardous obstacles. Traditional rescue operations in such conditions are typically slow, resource-intensive, and expose rescuers to significant risks, which limits the efficiency of emergency response. This paper presents the conceptual design of an AI-enabled UAV platform for rapid rescue in flood conditions, particularly in hazardous and inaccessible areas. The system integrates a custom, reconfigurable payload mechanism comprising four articulated arms mounted beneath the drone, each with six degrees of freedom driven by RC servomotors. This configuration affords high adaptability, enabling secure transport and active release of diverse rescue supplies (e.g., medical boxes, essential goods, life buoys). In parallel, a deep learning-based human detection model built on YOLOv12 is trained on aerial images captured in flood scenarios to rapidly detect and localize victims from the UAV perspective. The detection outputs provide immediate target cues, allowing the drone to quickly acquire the victim's location and deliver the payload directly within reach, thereby reducing the cognitive and physical burden on both the operator and the victim. Mechanical simulations of the four-arm mechanism across multiple payload types, together with training outcomes of the YOLOv12 detector, indicate the feasibility of the proposed approach and its suitability for time-critical field conditions. Overall, the combination of AI-assisted victim localization and a flexible, active payload mechanism reduces human intervention, improves delivery accuracy, and supports a smart, efficient, and readily deployable solution for disaster-response operations.