Lyrebird: Bringing DJI SDK-Compatible UAVs into the MAVLink Multi-Robot Ecosystem

Commercial off-the-shelf UAVs are attractive research platforms because they tend to be reliable and widely available. However, their closed and proprietary interfaces make them difficult to integrate into autonomous robotics experiments, where perception, coordination, and control are often built around open protocols and frameworks such as MAVLink, PX4, and ROS 2. 

We present Lyrebird, an open-source abstraction layer for DJI UAVs ranging from low-cost consumer models to enterprise platforms. Lyrebird translates standard interfaces into DJI-specific SDK operations, thereby presenting supported UAVs as MAVLink-compatible vehicles. Lyrebird runs on the Android device or remote controller connected to the UAV, allowing ground-station software to control the aircraft without directly interacting with DJI's SDK.

Through MAVLink 2, Lyrebird provides telemetry, flight control, vehicle configuration, autonomous waypoint missions, and camera and gimbal control. HTTP and TCP complement MAVLink for higher-throughput data transfer and functionalities that do not map naturally onto MAVLink messages. Lyrebird also provides a unified Python API, ROS 2 interfaces following PX4-style conventions, automatic discovery and identification of multiple UAVs on the same network, and live video streaming to external applications.

Lyrebird builds on WildBridge, an earlier version deployed for multi-UAV control in wildlife-monitoring and search-and-rescue scenarios. We characterise communication latency, telemetry and control performance, video streaming, and multi-UAV operation. Flight experiments include up to eight UAVs, with six DJI platforms integrated alongside two PX4-based drones and controlled through QGroundControl in Mission and Offboard modes.

By exposing DJI UAVs through standard robotics interfaces, Lyrebird reduces hardware integration effort and accelerates the transition from software development to real-world experiments. Control, coordination, and autonomy software can therefore be reused across single-, multi-, and heterogeneous-UAV systems with limited platform-specific adaptation. Software and documentation are openly available online (github.com/SDU-UAS-Center/lyrebird).