Toward Autonomous Ecological Monitoring with Betaflight FPV Drones: Test-Stand Validation of a Lightweight Visual Servoing Pipeline

Abstract

Small first-person-view (FPV) drones provide a flexible platform for ecological and environmental monitoring tasks, including the localization of pollution sources, inspection of contaminated areas, and low-disturbance observation of wildlife. However, Betaflight, one of the most widely used flight-control platforms for FPV drones, does not provide native support for autonomous vision-based target tracking. This study presents a lightweight onboard visual servoing framework that integrates real-time target tracking, image-based feedback control, and direct interaction with a Betaflight flight controller. Target tracking is performed using NanoTrack on a Raspberry Pi 5, while the target position estimated from the video stream is converted into a visual tracking error and processed by a PI controller to generate yaw-control commands. These commands are transmitted to the flight controller through the MSP Override mechanism, enabling closed-loop vision-based target centering without modifying the Betaflight firmware. A unified FlightController abstraction is introduced to support both simulation and physical deployment using the same high-level control architecture. The simulation environment combines ArduPilot SITL, Gazebo, and MAVLink, whereas the experimental platform employs Betaflight and the MSP protocol. The proposed framework is evaluated through a progressive validation procedure ranging from software-in-the-loop simulation to experiments with a physical FPV drone mounted on a yaw-only test stand. Experimental validation consists of two data-logged test-stand runs totaling 111 s, during which the system tracked a hand-moved visual target and commanded the drone to rotate in real time to keep it centered. The primary focus of this work is a proof-of-concept demonstration of end-to-end system integration, with the main contribution lying in the integration of existing perception, control, and communication components into a functional vision-to-command pipeline for this class of hardware. Closed-loop operation is demonstrated indoors using high-contrast visual targets and yaw-axis rotation. Therefore, the presented results should be interpreted as validation of the system architecture and integration concept rather than as a comprehensive evaluation of autonomous target tracking under field conditions. Quantitative evaluation using repeatable target trajectories, extension to free-flight operation, and field trials involving ecologically relevant targets constitute the next stage of this research. The proposed framework is intended as a computationally lightweight and hardware-compatible foundation for further development of vision-based target tracking and visual servoing on Betaflight-controlled FPV drones for ecological and environmental monitoring applications.