Autonomous aerial sensing enables wide-area terrain analysis, supporting damage assessment and target detection across large environments under changing light and weather conditions.
Our system integrates ArduPilot with a custom-built flight platform, combining a tailored airframe and flight controller with onboard perception capabilities. We developed a vision pipeline capable of detecting and classifying over 1000 objects, enabling the drone to dynamically adjust its motion and track targets in real time. Once aligned above a target, the system transitions to a stabilized hold mode for precise positioning. The platform is fully integrated with essential sensors, including GPS, and has been tested under real-world conditions to ensure reliability, robustness, and mission readiness for the international stage.
Our system leverages a YOLO-based vision model to detect and classify targets in real time, enabling the drone to make intelligent decisions during flight. Detection outputs are directly integrated with the flight controller, allowing dynamic adjustment of motion for precise tracking and positioning. Through tightly coupled communication between perception, control, and navigation subsystems, the platform operates as a unified autonomous system. This integration enables reliable performance in complex environments, supporting coordinated mission execution, environmental assessment, and target engagement across large operational areas.