GPS-Denied UAV-UGV Relative Positioning System via Vision-UWB Fusion

Locating drones in GPS-denied environments is a challenging task, particularly for hardware platforms with limited computational resources and sensors. A common approach for positioning a drone within the environment is to cooperate with an unmanned ground vehicle (UGV) by estimating the drone’s relative position on the UGV frame. In this paper, we propose a real-time drone positioning system that uses a Vision-UWB fusion framework. The fusion system involves two steps: 1) processing the state estimation using UWB ranging measurements and vision information separately, and 2) fusing the estimations from UWB and vision to provide the final state estimation for the drone. To evaluate the accuracy of our method, we deployed an accurate localization system, named OptiTrack, as the ground truth system. The experimental results demonstrate the accuracy and robustness of the developed system, which overcomes the limitations of individual sensors and enhances their advantages.