Nectar SDK: Open-source ROS 2 toolkit for autonomous drones

Black Bee Drones, an autonomous drone team at Brazil's Federal University of Itajubá (UNIFEI), has open-sourced Nectar SDK — a ROS 2 companion stack designed to simplify drone development for research and competition. The toolkit consolidates control, camera feeds, and perception into one unified interface, so developers don't have to rebuild mission-critical code for every new platform. It currently supports ArduPilot and PX4 autopilots through multiple transports, including MAVROS, direct MAVLink, and PX4's uXRCE-DDS, while also offering experimental support for Parrot Bebop and Crazyflie micro-drones.

The SDK is hardware-flexible: it runs on Jetson Orin Nano (JetPack 6, Humble) and Raspberry Pi 4/5, with Docker and CI coverage for ROS 2 Humble, Jazzy, and Kilted across amd64 and arm64. The vision stack handles USB, RealSense, OAK-D, Pi Camera, and ROS image topics, with built-in algorithms for ArUco, color, line, distance, optical flow, and MediaPipe. For AI, Nectar integrates Ultralytics YOLO, Hugging Face Transformers, and RF-DETR for detection, segmentation, and classification tasks, plus a nectar-ai CLI for training, evaluation, and prediction.

A standout feature is the DroneFactory pattern: a single API lets you switch between autopilot backends by changing a few lines of configuration — ideal for labs that test multiple platforms. The project also includes a Qt6 GUI and Gazebo SITL simulation for the same ArduPilot and PX4 backends. Nectar SDK has proven itself in high-pressure environments, earning 3rd place at IMAV indoor competitions in 2023 and 2025, and 2nd place at SAE Eletroquad 2026. The code is available on GitHub under Apache 2.0, inviting collaboration from other research teams and aerial robotics groups.