Accurate indoor positioning remains a significant challenge for first responders, primarily due to the absence of Global Navigation Satellite System (GNSS) signals and the inherent complexity of indoor environments, including multipath propagation, non-line-of-sight conditions, and frequent occlusions. Reliable localization is critical for situational awareness, team coordination, and responder safety during operations in unfamiliar buildings. Over recent years, micro-electro-mechanical systems (MEMS) inertial measurement units (IMUs) have become a key enabler for infrastructure-free pedestrian navigation. In particular, foot-mounted IMUs can exploit the natural gait cycle through zero-velocity updates (ZUPTs) during stance phases, reducing velocity errors and stabilizing attitude estimation. Nevertheless, even with ZUPTs, position errors cannot be fully bounded and residual drift accumulates over time, especially in long missions and under irregular motion patterns such as running or crawling.

To mitigate drift, prior work often fuses inertial navigation with external information such as floor-plan constraints, ultra-wideband (UWB) ranging to fixed anchors, wireless local area network (WLAN) or Bluetooth Low Energy (BLE) measurements, or fingerprinting-based radio maps. While these approaches can improve accuracy, they typically require prior knowledge (e.g., anchor coordinates, building plans) or pre-installed infrastructure, which may be unavailable, outdated, or impractical in emergency response scenarios. Infrastructure-less concepts such as on-the-fly anchor deployment reduce pre-survey requirements but still impose operational burden and require additional hardware to be placed in the environment.

In realistic deployments, responders typically enter buildings as teams and can exchange information when they come into proximity. This paper introduces a real-time cooperative localization framework based on tightly integrated, foot-mounted IMU and UWB units, where the UWB radio is co-located and time-synchronized with the foot-mounted IMU. Co-location simplifies calibration, provides consistent motion dynamics, and allows ranging epochs to align naturally with gait events (e.g., stance phases) where inertial uncertainty is reduced. The proposed method uses peer-to-peer UWB ranging primarily as inter-agent distance constraints, without requiring fixed anchors, floor plans, or radio maps.

We adopt a centralized cooperative architecture. On each responder, a local pedestrian dead reckoning (PDR) solution is computed from foot-mounted IMU data using an extended Kalman filter (EKF) with ZUPT pseudo-measurements. These local PDR estimates, together with intermittent inter-agent UWB range observations, are streamed to a central processing unit. At the central unit, a cooperative estimator refines the trajectories of all responders in a shared reference frame by fusing the PDR motion information with inter-agent distance constraints. The refined position estimates are then streamed back to the responders’ devices to support real-time operation.

We evaluate the approach using real measurements collected in a multi-floor office environment. The scenario comprises two independent responder groups, each consisting of two people, entering the building simultaneously and navigating across multiple floors largely independently. Distances are measured continuously within each group and opportunistically between groups whenever proximity permits (for example, when teams meet or pass nearby). The results demonstrate that cooperative inter-agent ranging, combined with the centralized fusion approach on top of foot-mounted EKF-based PDR, can substantially bound drift relative to standalone inertial navigation while maintaining fast deployment and minimal operational overhead.

Cooperative UAV Positioning in GNSS-Degraded Environments Using Frequency-Hopped UWB Networks

Reliable positioning is a fundamental requirement for cooperative unmanned aerial vehicle (UAV) operations. In urban canyons and forested environments, Global Navigation Satellite System (GNSS) availability often varies spatially within a single formation due to signal obstruction, multipath propagation, or intentional jamming. Some platforms may maintain valid satellite solutions while others experience partial degradation or complete denial. This paper presents a cooperative positioning framework based on ultra-wideband (UWB) ranging that enables robust localization under such heterogeneous conditions.

Each UAV is equipped with a Qorvo DW1000 UWB transceiver and performs two-way ranging with both fixed ground anchors of known position and neighboring UAVs. Ground anchors provide an absolute reference when GNSS is unavailable, while inter-UAV ranges introduce cooperative geometric constraints that allow GNSS-capable platforms to support GNSS-denied neighbors. State estimation is performed locally on each platform using a distributed Extended Kalman Filter, fusing inertial measurements, barometric altitude, UWB ranges, and available GNSS observations without reliance on centralized processing.

To improve robustness against interference, the proposed system employs adaptive frequency hopping across the DW1000-supported IEEE 802.15.4 UWB channels with approximately 500 MHz instantaneous bandwidth in the 3.5 - 6.5 GHz range. Randomized channel switching is performed between consecutive ranging sessions, while individual two-way ranging exchanges remain on a single channel to preserve time-of-flight measurement integrity. This inter-session hopping strategy mitigates narrowband interference and increases resilience to intentional jamming without modifying the underlying ranging protocol.

Scalability challenges arising from the quadratic growth of pairwise ranging links are addressed through a time-division multiple access scheme optimized for aerial network geometry, with link prioritization based on GNSS availability and cooperative observability. Experimental results obtained with representative multi-UAV flight patterns demonstrate horizontal positioning errors of 15-20 cm for GNSS-denied platforms when cooperating with at least two GNSS-capable neighbors within a 100 m range. Frequency hopping provides a measurable improvement in interference tolerance compared to fixed-channel operation under identical conditions.

The proposed framework enables resilient cooperative positioning for UAV missions where GNSS reliability varies spatially or temporally, including emergency response and inspection tasks.

Achieving reliable navigation in dense urban or GNSS-denied environments remains a central challenge for autonomous aerial vehicles. For our application (e.g., infrastructure monitoring, law enforcement, search and rescue), we consider a heterogeneous swarm composed of high-flyers (HF) and low-flyers (LF). Whereas the HFs maintain high-quality GNSS navigation, the LFs operate in the urban canyons with degraded or denied GNSS-based navigation performance. Our proposed method relies on Ultra-Wide-Band inter-vehicle ranging to form geometric constraints; HF agents act as mobile navigation anchors, while LF agents obtain their global state through cooperative multilateration. Future implementations may include vision- or LiDAR-based map matching, but they are not considered in the current work as they require the participant to have a prior map or build one.
Building on recent work in UWB-aided factor-graph localization for small UAS, this paper presents a decentralized factor-graph–based cooperative localization framework. Each vehicle infers its trajectory by maintaining and solving a local sliding-window factor graph (i.e. factor graph optimization, FGO) incorporating UWB ranges, inertial measurements, and motion models to . Global observability (i.e., the trajectory estimates are anchored to a global reference frame) through the incorporation of the HFs’ absolute position estimates into the LF factor graphs as prior factors. However, sharing priors may introduce statistical dependencies among the LF estimates as the underlying HF states are common information across the network. To preserve estimator consistency under these unknown cross-estimator correlations, we employ methods similar to covariance intersection (CI) within the factor-graph fusion process. Typically, CI provides a mathematically conservative fusion rule that guarantees bounded covariance growth and prevents violation of probabilistic consistency without requiring explicit cross-covariance terms. This ensures that (multi-hop) information propagation does not lead to overconfident or biased state estimates, even under asynchronous communication and partial measurement availability.
The paper will discuss, in detail, the proposed decentralized cooperative FGO method and show the results from its evaluation in simulation and its validation through multi-UAV flight experiments. Results demonstrate that (i) decentralized factor-graph estimation with CI maintains estimator consistency even when using correlated information, (ii) UWB-based geometric constraints remain stable in cluttered or partially occluded environments, and (iii) the overall framework enables robust navigation without prior maps, even in complex urban-like settings.

Navigation in challenging GNSS-denied environments for emergency forces requires high quality PNT (positioning, navigation and timing) solutions. For indoor scenarios in buildings or tunnels, several sensors have been proven to be suitable. Ultra-Wideband (UWB) is currently the contender for becoming the standard ranging and positioning method for such scenarios. It offers decimeter-level accuracy in optimal cases and ranging up to 100 m. However, UWB requires infrastructure , and multipath hampers its reliability. UWB-based positioning of pedestrians is therefore sometimes augmented with step information deduced from IMUs (Inertial Measurement Units). This paper presents an additional improvement by introducing a mobile robot, which accompanies the pedestrian. The robot itself uses a LiDAR (Light Detection and Ranging),wheel odometry, and UWB to estimate its position. Using a collaborative algorithm, the pedestrian’s location is enhanced by incorporating UWB ranges between the robot and the pedestrian. This human-machine teaming improves reliability and accuracy due to the additional sensors on the robot, which serves as a moving UWB anchor for additional ranges. The paper is structured into three sections: first, the developed UWB ranging and communication network is explained. It features a custom TDMA (Time Division Multiple Access) scheme on the UWB hardware. Then, the centralized factor graph optimization (FGO) with all measurements (UWB, IMU, Robot-SLAM (Simultaneous Localization and Mapping)) is presented. Using the FGO architecture, inter-agent ranges and position changes can easily be integrated as factors. Finally, results from test measurements in indoor environments are presented. The advantages of the collaborative approach are evaluated and errors analyzed. The focus lies on the challenges of localization outside of a known UWB network, where geometry degrades or UWB anchors are not visible anymore, just by using ranges between agents and relative sensors.