U-VIO: Tightly Coupled UWB Visual Inertial Odometry for Robust Localization

Abstract

The conventional visual-inertial odometry (VIO)-based localization techniques perform well in environments where stable features are guaranteed. However, their performance is not assured in poor feature quality and quantity conditions. As a solution to this, the U-VIO, a tightly coupled UWB visual-inertial odometry algorithm, is proposed in this paper. In the front-end, distance measurement between a static UWB anchor and a keyframe was considered as a residual for pose estimation. In the back-end, after finding the loop closure relationship with the current keyframe among the previous keyframes based on visual information, UWB loop constraints were added. The proposed algorithm was evaluated with data collected using a UGV (Unmanned Ground Vehicle). In the experimental analysis, the case where the UWB factor was added only to the front-end and the case where the back-end was additionally considered were compared. The proposed algorithm that closely uses UWB factors for the entire graph structure showed the most robust pose estimation performance through these evaluations.