ULOC: Learning to Localize in Complex Large-Scale Environments with Ultra-Wideband Ranges

TL;DR

A learning-based framework named ULOC for Ultra-Wideband (UWB) based localization in complex, large-scale environments based on MAMBA that learns the ranging patterns of UWBs over a complex, large-scale environment is proposed.

Abstract

While UWB-based methods can achieve high localization accuracy in small-scale areas, their accuracy and reliability are significantly challenged in large-scale environments. In this paper, we propose a learning-based framework named ULOC for Ultra-Wideband (UWB) based localization in such complex large-scale environments. First, anchors are deployed in the environment without knowledge of their actual position. Then, UWB observations are collected when the vehicle travels in the environment. At the same time, map-consistent pose estimates are developed from registering (onboard self-localization) data with the prior map to provide the training labels. We then propose a network based on MAMBA that learns the ranging patterns of UWBs over a complex large-scale environment. The experiment demonstrates that our solution can ensure high localization accuracy on a large scale compared to the state-of-the-art. We release our source code to benefit the community at https://github.com/brytsknguyen/uloc.

Figures (10)

• Figure 1: Large scale environment with ten anchors (red squares) deployed over a 400m $\times$ 200m campus area. The path of the vehicle is marked in orange. On the bottom two plots, we show the UWB measurements to the anchors during the travel, and number of anchors seen over time on a 1s sliding window (bottom). We can see that there is hardly any location where measurements to all ten anchors are available. Moreover, due to multi-path and NLOS conditions, the outliers and observation losses are also quite frequent. To the best of our knowledge, this environment being investigated in this paper is the largest with high-accuracy ground truth among existing works on UWB-based localization.
• Figure 2: A simple UWB-based localization scheme in small-scale environments. If we define ${\bm{\mathbf{R}}}, {\bm{\mathbf{p}}}, d, \bm{a}, \bm{b}, a, b$ respectively as the robot orientation, its position, the range, anchor position, UWB tag's offset, and some scaling factors, the ranging model $d = a\left\lVert{\bm{\mathbf{p}}} + {\bm{\mathbf{R}}}\bm{b} - \bm{a}\right\rVert + b$ can be used for full pose localization of UWB in small-scale environments. For the estimate to converge during a certain period of time, at least range measurements to each anchor have to be obtained.
• Figure 3: The data collection setup.
• Figure 4: The subtle difference between range measurements by two UWB tags can contain information on the orientation ${\bm{\mathbf{R}}}$ as explained in Fig. \ref{['fig: small scale environment']}.
• Figure 5: Top-left, a survey-grade prior map (SGPM, in RGB) compared with a loop-closure-enabled prior map from SLICT nguyen2023slict (in red). We find that the SLICT-built prior map deviates from the SGPM at certain locations, especially in the z-direction, as shown in the bottom-left image.