Modular Meshed Ultra-Wideband Aided Inertial Navigation with Robust Anchor Calibration

TL;DR

This work addresses scalable, drift-resistant localization in GPS-denied environments by introducing a modular, filter-based fusion framework that supports fully meshed UWB ranging among a time-varying set of devices. It combines cross-covariance factorization with two decoupling strategies (DP and DAH) to enable true modularity and sensor hot-swapping, while extending anchor calibration to multiple tags and anchors through a coarse-to-fine calibration pipeline that includes nonlinear refinement and RANSAC-based outlier rejection. Key contributions include (i) an online anchor-position and pairwise-bias calibration extending UVIO-style methods to multi-tag/anchor setups, (ii) a fully meshed SDS-TWR ranging scheme with a scalable scheduling protocol, and (iii) a real-time capable, modular estimator validated on real flights with a public dataset and toolchain. The results demonstrate accurate anchor initialization (average errors around ${ ilde{f p}}_{ ext{A}} o 0.22$ m in offline tests), robust online self-calibration, and real-time operation on a UAV, highlighting the practical impact for indoor navigation and sensor rehab/augmentation scenarios.

Abstract

This paper introduces a generic filter-based state estimation framework that supports two state-decoupling strategies based on cross-covariance factorization. These strategies reduce the computational complexity and inherently support true modularity -- a perquisite for handling and processing meshed range measurements among a time-varying set of devices. In order to utilize these measurements in the estimation framework, positions of newly detected stationary devices (anchors) and the pairwise biases between the ranging devices are required. In this work an autonomous calibration procedure for new anchors is presented, that utilizes range measurements from multiple tags as well as already known anchors. To improve the robustness, an outlier rejection method is introduced. After the calibration is performed, the sensor fusion framework obtains initial beliefs of the anchor positions and dictionaries of pairwise biases, in order to fuse range measurements obtained from new anchors tightly-coupled. The effectiveness of the filter and calibration framework has been validated through evaluations on a recorded dataset and real-world experiments.

Figures (12)

• Figure 1: Spatial frame constellation of the proposed modular aided inertial navigation framework. A calibrated UWB ranging infrastructure is extended at runtime by additional stationary ranging devices. After a coarse initialization, these new anchors positions (purple) are estimated in filter framework.
• Figure 2: Block diagram of the proposed sensor fusion framework utilizing the DAH or DP approach proposed in jung_modular_2021, consisting of an instance handler ${\mathsf{H}}$ maintaining instances (nodes) of a specific type.
• Figure 3: Congestion free fully-meshed scheduling protocol. Each node listens and waits for it's preceding node to finish the one-to-all ranging cycle. Four arrows indicate a full SDS-TWR cycle consisting of four messages, which needs to be completed within $\Delta t$.
• Figure 4: Left image shows the body trajectory (black line), the stationary anchor ${{}^{{\cG}}_{{}}{{\vp}}^{}_{{\cA}}} = [10;10;10]$ (blue circle), intermediate results (red circle) using random samples of the recorded measurements shown in the right images. The measurements contained [15]% outliers with a ranging and tag position standard deviation of [0.1]m. The yellow line shows the fitted model, after removing outliers.
• Figure 5: Shows a subset of the deployed UWB ranging devices in a cluttered environment, covered by a motion capture system. In total nine anchors consisting of a RPi4 companion computer and a Qorvo MDEK1001 UWB transceiver are used. The UAV is equipped with two ranging devices, a RPi4, a Pixhawk 4 flight controller, and reflective markers for ground truthing.