Infrastructure-less UWB-based Active Relative Localization

TL;DR

This work tackles infrastructure-less relative localization for multi-robot systems using only UWB distance measurements, removing the restriction of static anchor platforms. It combines a geometry-aware anchor placement (isosceles configuration), a novel loss that fuses GDOP with a short-range measurement model, and a SAC-based DRL controller to move the anchor-equipped robot to minimize the TagBot's position error, achieving up to 60% improvement over prior methods. The approach is validated through extensive simulations and real-world experiments, demonstrating robust performance and generalization across static and dynamic TagBot scenarios. The results highlight the practical potential of active, UWB-based localization in non-line-of-sight environments and lay groundwork for future sensor fusion and aerial/mobile deployments.

Abstract

In multi-robot systems, relative localization between platforms plays a crucial role in many tasks, such as leader following, target tracking, or cooperative maneuvering. State of the Art (SotA) approaches either rely on infrastructure-based or on infrastructure-less setups. The former typically achieve high localization accuracy but require fixed external structures. The latter provide more flexibility, however, most of the works use cameras or lidars that require Line-of-Sight (LoS) to operate. Ultra Wide Band (UWB) devices are emerging as a viable alternative to build infrastructure-less solutions that do not require LoS. These approaches directly deploy the UWB sensors on the robots. However, they require that at least one of the platforms is static, limiting the advantages of an infrastructure-less setup. In this work, we remove this constraint and introduce an active method for infrastructure-less relative localization. Our approach allows the robot to adapt its position to minimize the relative localization error of the other platform. To this aim, we first design a specialized anchor placement for the active localization task. Then, we propose a novel UWB Relative Localization Loss that adapts the Geometric Dilution Of Precision metric to the infrastructure-less scenario. Lastly, we leverage this loss function to train an active Deep Reinforcement Learning-based controller for UWB relative localization. An extensive simulation campaign and real-world experiments validate our method, showing up to a 60% reduction of the localization error compared to current SotA approaches.

Figures (5)

• Figure 1: Overview of the proposed approach. In a UWB-based relative localization context, the performance is affected by the geometrical distribution of the sensors in the environment. Contrary to the SotA approaches (bottom figure with red bar), we develop an active method (top figure with green bar) that exploits the movement of the robot to enhance the localization performance.
• Figure 2: Configurations from 1 to 4. On the left: simulated GDOP (continuous lines) and analytical GDOP (dashed lines) calculated on 3 circumferences of radius 50 cm (blue), 74 cm (light blue), and 1 m (red), with respect to the AnchorBot. On the right: analytical GDOP on a ring of points around the AnchorBot.
• Figure 3: Comparison between (a) the GDOP and (b) the UWB Relative Localization Loss. The grey dots are the anchors, while the dashed green line represents the footprint of the AnchorBot. Cold colors indicate low values, and warm ones refer to high values.
• Figure 4: Plots of the dynamic experiments for both SUL-EQ (top) and AUL-IS (bottom) are shown for the Straight Line and Circle experiments.
• Figure 5: Circle dynamic experiment with a radius of 85 cm in real-world. The colored outlines in the figures indicate the GDOP value for the considered timestamp according to the colorbar scale.