Range-Focused Fusion of Camera-IMU-UWB for Accurate and Drift-Reduced Localization

Abstract

In this work, we present a tightly-coupled fusion scheme of a monocular camera, a 6-DoF IMU, and a single unknown Ultra-wideband (UWB) anchor to achieve accurate and drift-reduced localization. Specifically, this letter focuses on incorporating the UWB sensor into an existing state-of-the-art visual-inertial system. Previous works toward this goal use a single nearest UWB range data to update robot positions in the sliding window (“position-focused”) and have demonstrated encouraging results. However, these approaches ignore 1) the time-offset between UWB and camera sensors, and 2) all other ranges between two consecutive keyframes. Our approach shifts the perspective to the UWB measurements (“range-focused”) by leveraging the propagated information readily available from the visual-inertial odometry pipeline. This allows the UWB data to be used in a more effective manner: the time-offset of each range data is addressed and all available measurements can be utilized. Experimental results show that the proposed method consistently outperforms previous methods in both estimating the anchor position and reducing the drift in long-term trajectories.