Accurately Tracking Relative Positions of Moving Trackers based on UWB Ranging and Inertial Sensing without Anchors

TL;DR

This work introduces a multi-stage filtering pipeline through which their system estimates the relative layout of all tracking nodes within the group, integrating a custom Extended Kalman filter with a refinement step via multidimensional scaling (MDS).

Abstract

We present a tracking system for relative positioning that can operate on entirely moving tracking nodes without the need for stationary anchors. Each node embeds a 9-DOF magnetic and inertial measurement unit and a single-antenna ultra-wideband radio. We introduce a multi-stage filtering pipeline through which our system estimates the relative layout of all tracking nodes within the group. The key novelty of our method is the integration of a custom Extended Kalman filter (EKF) with a refinement step via multidimensional scaling (MDS). Our method integrates the MDS output back into the EKF, thereby creating a dynamic feedback loop for more robust estimates. We complement our method with UWB ranging protocol that we designed to allow tracking nodes to opportunistically join and leave the group. In our evaluation with constantly moving nodes, our system estimated relative positions with an error of 10.2cm (in 2D) and 21.7cm (in 3D), including obstacles that occluded the line of sight between tracking nodes. Our approach requires no external infrastructure, making it particularly suitable for operation in environments where stationary setups are impractical.

Figures (8)

• Figure 1: Our tracking system estimates relative positions among a moving group of tracking nodes, demonstrated here for RC cars and pose tracking of a moving person. Each tracker has a single-antenna UWB radio, a 6-DoF inertial sensor, and a 3-DoF magnetometer for state estimation.
• Figure 2: Overview of our tracking system. Each tracking node estimates its own 3D orientation state through an embedded VQF filter. Through UWB broadcasts, here at $t_i$ and $t_j$, all nodes estimate pairwise distances, which serve as observations in the update step of our Extended Kalman Filter to estimate relative positions.
• Figure 3: Our pipeline includes individual Extended Kalman Filters for each tracking node's orientation as well as EKFs for each pair of nodes for relative positions. A key part of our pipeline is the feedback loop that stabilizes individual EKF instances with the output from MDS to achieve spatial consistency.
• Figure 4: Example of a ranging transaction with two responders. The timestamps required to resolve time-of-flight are included in the UWB message payload and thus broadcasted to all participants in the constellation.
• Figure 5: Each embedded tracking node features a base micro-controller PCB with inertial and magnetics sensing, with an extension for UWB ranging.