Micro unmanned aerial vehicles (micro-UAVs) are small enough to reach confined spaces that larger robots cannot access, but too small to carry the sensing and computing power required for autonomous flight. We move the localization stack entirely off the aerial platform onto a quadruped robot with a 7-degree-of-freedom (DOF) arm, which supplies the micro-UAV (27 g bare, 42 g with fiducial markers) its full 6-DOF pose. A camera at the arm's end-effector detects AprilTag fiducial markers on the drone and composes that observation with the quadruped's own self-localization to place the drone in a shared map frame, so the ground robot localizes its partner, rather than only tracking it relative to the camera. The arm acts as an actively-controlled observer, repositioning to keep the drone in view as both robots move; the drone carries only an inertial measurement unit and fuses the external pose to fly commanded setpoints. In lab flights the external pose is accurate to 12-16 mm, enough to fly the drone autonomously within 2-5 cm of motion-capture-fed control. Having the quadruped actively follow the drone reduces the tracking error from 11.0 cm to 6.9 cm by holding the camera in the close range, where the markers are most accurate.
Figures & tables
Figure 1: The integrated hardware. The Boston Dynamics Spot carries the Spot Arm (7 DOF including the gripper)—whose end-effector camera supplies the AprilTag observations of the drone that drive the perception path—and the Jetson AGX Orin on its payload rails. The Crazyflie 2.1+ with its four-AprilTag chassis is the aerial platform the arm camera observes.
Figure 2: System architecture. A perception path (the quadruped’s sensors and the arm-mounted camera through the quadruped’s SLAM, the apriltag_relocalizer , and AprilTag drone-pose estimation) and a command path (trajectory commander through the micro-UAV control stack and radio link to the drone) run concurrently in a single ROS 2 domain; control consumes the composed Tmap→drone and the drone publishes no pose of its own.
Figure 3: Experiment 1: AprilTag vs. MOCAP position and orientation error by camera resolution (raw data and Horn/Markley aligned).
Figure 4: Experiment 2: MOCAP vs AprilTag-controlled flight. (a) Commanded trajectory (dashed) and per-waypoint median flown position with IQR; AprilTag overshoots the y = ±0.5 turn-arounds. (b) Age of the control pose reaching the drone: AprilTag is a median 72 ms stale vs 2 ms for MOCAP, explaining the overshoot.
Figure 5: Experiment 3: Planned trajectory tracking error by quadruped motion using 640 × 480 resolution. (a) tracking error boxplots per quadruped motion condition; (b) tracking error over time per quadruped motion condition
1 – School of Electrical and Data Engineering, University of Technology Sydney, New South Wales, Australia · 2 – Institute of Aerospace Engineering and Technology, Duy Tan University, Da Nang, Vietnam · 5 – Department of Artificial Intelligence and Robotics, Sejong University, Seoul, South Korea +2