Future Martian exploration will require autonomous, high-integrity Positioning, Navigation, and Timing (PNT) capabilities that do not rely on Earth-based Global Navigation Satellite Systems (GNSS). This work presents an approach for adapting a software-defined receiver to a conceptual Mars satellite navigation system, implemented by adapting the open-source GNSS-SDR framework and driven by a (Global Positioning System) GPS-like signal simulated for Mars-centric orbital and physical conditions.
A minimal-variation approach is chosen to IS-GPS-200 message format, bit allocations and subframe layout, while regenerating the contained parameter values to accurately reflect the Martian dynamic environment to produce broadcast ephemerides suitable for medium-altitude Mars orbits. Keplerian elements and their derivatives are fitted from propagated Mars orbits over multi-hour intervals to ensure the broadcast message is fully compliant with terrestrial GNSS scaling and data structures.
The simulation environment generates signal I/Q files from a configurable Mars constellation in the 3000–8000 km altitude range within a Mars-Centred Mars-Fixed (MCMF) frame. Thes I/Q files will be processed by a modified GNSS-SDR receiver that incorporates Martian models, removes Earth atmospheric models and replaces Earth gravity constants on the navigation solution.
The ongoing work integrates the simulated Mars-adapted signals with a modified GNSS-SDR receiver chain. The next development steps will focus on validating full acquisition, tracking, navigation message decoding, and Mars-centric PVT (Position, Velocity, and Timing) capabilities using the regenerated broadcast parameters. The proposed simulation–receiver architecture is expected to provide a reproducible foundation for analysing Mars constellation geometries, evaluating PNT availability, and assessing the achievable performance of future Mars PNT services. This framework will serve as a baseline for subsequent research on robust signal design, receiver algorithms, and planetary-scale navigation infrastructure.

Low-cost GNSS receivers onboard nanosatellites often operate with reduced measurement sampling rates due to power, data volume, or onboard processing constraints. This research investigates the impact of sparse GNSS measurement intervals on precise orbit determination (POD) of low Earth orbit (LEO) satellites using raPPPid software. GNSS observations from an Astrocast nanosatellite are analyzed for measurement intervals ranging from the nominal 0.5 seconds up to 5, 10, 15, and 20 seconds. Two POD strategies are evaluated based on different mission requirements. First is high-rate dynamic propagation with state prediction at the nominal 0.5 seconds rate, while applying measurement updates at reduced intervals. Second is reduced-rate model propagation and correction, where state prediction and measurement updates are both performed at the sparse sampling interval.

Results show that increasing the measurement interval primarily affects the early filter convergence phase, with reduced observability leading to temporary orbit instability. After convergence, orbit solutions obtained with sparse measurements show similar long-term behavior to the nominal case, with only moderate degradation in 3D position and velocity accuracy, particularly when high-rate dynamic propagation is maintained. However, sparse sampling significantly impacts cycle-slip detectability, especially for Doppler-based methods. Cycle slips may accumulate over extended intervals or trigger false detections under high dynamics and low elevation angle conditions. The study discusses Doppler–phase consistency metrics adapted for variable sampling intervals and quantifies the trade-offs between measurement rate, propagation strategy, detection robustness, and final POD accuracy. The results provide practical guidance for nanosatellite missions operating under sparse GNSS sampling constraints.

The EGNSS MATE project partners SBB, DLR and IABG have researched the topic of safe train localisation within the European Rail Traffic Management System context. This paper highlights the contributions to jamming and spoofing testing and the evaluation of the new Galileo features for their use in the railway environment.
In the EGNSS MATE project, a series of dedicated test analyses were conducted to investigate the susceptibility of railway PVT GNSS-based systems to radio frequency interferences—particularly intentional threats such as jamming and spoofing. This was performed to understand the impact of such events on GNSS receivers as future evolutions of the ERTMS/ETCS foreseen a more prominent role of GNSS systems concerning safe PVT navigation solutions. In the project, a set of jamming and spoofing tests were defined and conducted in a dedicated test campaign. In addition, within the project, the new Galileo services, that are the Galileo High-Accuracy-Service (HAS) and OS-NMA (Open Service – Navigation Message Authentication) were evaluated for their usage in rail applications. It was found out that there is a small improvement in terms of position accuracy performance by applying HAS correction data on GNSS measurements. Additionally, the OS-NMA detection capability under a live-spoofing event was demonstrated, showing that this service could deliver a real benefit for safety and security-relevant rail applications. In this communication, we report the main results of the GNSS test campaign.

Cold-atom interferometry (CAI) based inertial sensors present strong potential for future space applications, attributed to their stable long-term and precise measurements independent of external signals. This makes them beneficial for deep space missions where the Global Navigation Satellite System (GNSS) space service volume is either inaccessible or where GNSS signals suffer significant attenuation. However, the limitations of CAI quantum sensors, such as low measurement rates and ambiguities due to the measurement principle, necessitate the hybridization of quantum sensors with classical inertial sensors with higher bandwidth. The hybridization approach (Tennstedt et al., 2023) in this study follows two main objectives: First, data from high-rate sensors is used to solve the fringe ambiguity inherent to the sinusoidal nature of the raw CAI quantum sensor signal. Second, the superior long-term stability of the CAI quantum sensor principally enables real-time calibration of the classical inertial sensors. Previous research has shown that performance in terrestrial applications can be significantly enhanced depending on sensor parameters and trajectory dynamics (Gersemann et al., 2025), while future satellite gravity missions may also benefit from similar improvements. The CAI sensor hybridisation increases performance by one to two orders of magnitude in both applications. Currently, missions such as LuGRE and Artemis are pioneering a new paradigm for lunar navigation by using weak Global Navigation Satellite System (GNSS) signals. While these approaches provide a foundational framework, they achieve a moderate level of precision. Internal analysis conducted at IfE, consistent with the LuGRE mission’s concept confirms that geometric dilution of precision (GDOP), receiver sensitivity, and antenna gain are critical determinants of navigation accuracy in weak GNSS environments. The simulation results of hybrid classic quantum inertial navigation will be used to quantify potential performance gains relative to the navigation accuracy achieved by lunar missions like LuGRE and Artemis, offering an assessment of the proposed approach for lunar navigation.
To exploit the advantages of hybridization, the angular rates and specific forces recorded by the IMU are employed to predict the CAI phase shift. This data is subsequently integrated with the CAI observation through the application of an Extended Kalman Filter. The final phase shift is determined by resolving the fringe ambiguity of the CAI sensor, from which bias estimates intrinsic to classical inertial sensors are derived. These bias-corrected IMU data values are then utilized to compute the rover’s navigation states (position, velocity, and orientation). The potentials, fundamental limitations, and optimal interrogation durations of the CAI sensor will be analysed through typical trajectories for lunar rover navigation.
This work is funded by the Federal Ministry for Economic Affairs and Climate Action (BMWK), Projects 50NA2310A and 50NA2106.

Key words: Cold-atom interferometry (CAI), GNSS, Lunar Rover, Lunar navigation, Hybridisation

Global Navigation Satellite Systems (GNSS) are affected by various signal impairments in practical environments, including multipath, interference, and non-Gaussian measurement noise. This significantly reduces the reliability and positioning accuracy. One widely used filtering approach, as a solution, is the Kalman filter. However, these filters rely on the assumption of Gaussian noise and perform well only under such conditions. In recent years, a new filtering method based on the maximum correntropy criterion (MCC) has been explored as an alternative to the minimum mean square error (MMSE) criterion, which is the basis of traditional Kalman filters. The performance of MCC-based filters strongly depends on the choice of the kernel function. For instance, the Gaussian kernel MCC, which is conventionally used, tends to become unstable in multidimensional noise environments, and its performance heavily relies on precise bandwidth tuning. To address these limitations, this study introduces a Cauchy kernel-based maximum correntropy Extended Kalman Filter (CKMCC-EKF). The heavy-tailed and polynomial decay properties of the Cauchy kernel provide high robustness and substantially reduce sensitivity to bandwidth. To demonstrate the effectiveness of the proposed method, simulation experiments were conducted and compared with the traditional Extended Kalman Filter, the Gaussian-kernel MCC, and multi-kernel MCC filters. The results showed up to a 20% reduction in RMSE, highlighting the robustness of the proposed approach and its applicability to real-world GNSS scenarios.

Marine biofouling poses a significant challenge to maritime operations by increasing hull and propeller surface roughness, which elevates hydrodynamic resistance and fuel consumption. Depending on the severity, fouling can increase fuel use by up to 30%, leading to higher operational costs and elevated greenhouse gas emissions. Regular cleaning of hulls and propellers is therefore a critical maintenance activity, yet accurately quantifying its effectiveness remains difficult due to confounding operational and environmental factors.
Existing evaluation methods often rely on weakly informed analytics or conventional regression and machine‑learning models. While such data‑driven models can capture correlations, they struggle with predicting future performance, especially when operational conditions change. This predictive limitation arises because these models lack physical grounding and cannot reliably extrapolate beyond the data they were trained on. As a result, they are unable to isolate fouling effects or forecast the benefits of cleaning interventions with sufficient confidence.
To address this challenge, we introduce a hybrid modelling approach that combines physically informed RANS‑based surrogate models with machine learning. Rather than training directly on operational data—which is noisy, incomplete, and not well suited for long‑horizon prediction—our method focuses on learning the residuals between measured performance and RANS surrogate predictions. By letting the physics-based model provide the baseline prediction and tasking machine learning only with modeling the unexplained deviations, the approach preserves physical consistency while capturing data‑driven patterns related to fouling and operational variability. This residual‑learning strategy enables a fuel‑consumption prediction accuracy of approximately 2%, making forward‑looking assessments and optimal cleaning‑time decisions significantly more reliable.
We validate the methodology using performance data from an in‑water hull and propeller cleaning of the SNSPOOL Dina Scout conducted by Fleet Robotics. We first demonstrate that weakly informed and purely data‑driven analytics cannot reliably assess cleaning effectiveness. In contrast, our hybrid residual model clearly attributes 5–7% fuel savings in the three months following the maintenance event. These results show that hybrid physics‑informed, data‑driven modelling provides a robust foundation for quantifying—and justifying—the economic and environmental benefits of hull and propeller cleaning.

In recent years, Global Navigation Satellite System (GNSS) services have undergone rapid expansion, enabling their integration into increasingly complex and safety-critical applications. This growing dependence has intensified concerns regarding GNSS robustness and reliability, particularly in light of evolving geopolitical conditions that have led to a more contested radio-frequency environment. As a result, intentional interference in the form of jamming and spoofing is becoming more frequent and operationally relevant.

To address these threats, the Galileo GNSS has introduced Navigation Message Authentication (NMA) as a first line of defence against signal manipulation. However, advanced attack techniques-such as Secure Code Estimation and Replay (SCER) spoofing-, can circumvent NMA-based protections, highlighting the need for additional and complementary mitigation mechanisms.

This paper presents a family of GNSS resilience solutions developed by DLR GfR, designed to enhance protection against sophisticated spoofing scenarios beyond standalone authentication. Furthermore, it reports and discusses results from several studies and activities conducted in cooperation with the European Space Agency, demonstrating the performance, limitations, and operational applicability of these approaches. The findings contribute to ongoing efforts to strengthen GNSS resilience in European and global navigation infrastructures.

With continuously growing spectrum usage, the competition for bandwidth is also increasing.
To meet today‘s needs, new approaches and modification of existing radio infrastructure needs to be considered. One approach is to add synchronized aiding carriers to an existing medium frequency transmitter, which allows time-of-arrival estimation on the receiving phase. If a sufficient number of stations are received,
a positioning solution can be provided.
The resulting time of arrival as well as the calculated position can be used as backup and integrity test for another navigation solution such as global navigation satellite systems. This increases the resilience of positioning of any asset.
However, the existing transmitting antenna systems are not optimized for continuous stable phase transmission, which is important for achieving high positioning performance. This paper presents an optimization approach for existing MF antennas, combining analytical modeling and field verification to enhance phase transmission stability. Our study demonstrates the feasibility of this hybrid approach by analyzing the impact of antenna modifications on positioning accuracy using numerical simulations and experimental testing at the transmitter Groß Mohrdorf, Baltic Sea, Germany. We observed significant improvements in phase transmission stability, leading to improved positioning accuracy.
The results of this study have implications for navigation systems operating on medium-frequency bands, providing a promising solution for enhancing navigation performance.

In urban environments, absolute pose estimation based solely on GNSS is often degraded by multipath effects, resulting in poor accuracy. GNSS is also vulnerable to jamming and spoofing. Yet, many safety-critical navigation tasks demand highly accurate six degrees of freedom (6-DoF) pose estimates. The EGENIOUSS framework addresses these challenges by fusing data from inertial measurement units (IMUs), cameras via visual localisation (VL) and visual odometry (VO), and GNSS receivers from smartphone and low-cost mobile platforms. The EGENIOUSS system is therefore a complementary, redundant navigation framework.

Within EGENIOUSS, two independent VL components are implemented. The first VL module employs 3D meshes to estimate a 6-DoF pose, which is then fed into the main sensor-fusion pipeline together with GNSS and VO outputs. The second VL module uses City Geography Markup Language (CityGML) models as a reference data source and is implemented as an external verification step. The aim of the object-based VL component is not, as is usual in VL, to estimate the pose on the basis of a query image and its correspondence with the reference data. The object-based VL component is used to verify the pose from the sensor fusion based on the matching quality between query image and reference data. For this purpose, the CityGML models are rendered on the basis of a given pose. Due to the low memory requirements of CityGML models, the models can either be loaded directly on the platform and thus also used offline or loaded from the EGENIOUSS cloud solution. The rendered image is then used for matching with the query image. CityGML models are a challenging reference data type for VL due to their textureless, colourless and low-detail nature. Therefore, classical VL approaches and feature extractors and matchers often fail to match the query image with the rendered scene of the CityGML model. We have therefore implemented line extractors and a geometric line matching method.

In this contribution, we investigate the use of CityGML models as an independent reference data source from the mesh-based VL component in EGENIOUSS as a verification module for the pose from sensor fusion. We will evaluate the reliability of the verification module by testing different pose accuracies as input poses. We will investigate which matching thresholds indicate good, medium and bad input poses in order to develop a reliable verification module. This verification module provides the EGENIOUSS system with a further redundant, complementary safety layer that is independent in terms of methodology and data.

The EGENIOUSS project aims to develop a highly accurate and cost-effective positioning, navigation and timing system based on the integration of multiple geodata sources. These include visual localisation, in-situ sensor data and satellite-based augmentation services (Galileo High-Accuracy Service). The service is designed to support for smartphone-based use cases, including professional surveying in urban environments (QField app) and bicycle navigation with associated trip data collection (Naviki app). EGENIOUSS provides significant advantages, especially in areas where positioning quality is degraded due to building-induced shading and signal obstruction.

During the EGENIOUSS project, the EGENIOUSS service was implemented and tested within existing software packages in order to demonstrate its applicability, functional performance as well as the usability within real-world application contexts. The smartphone-based use cases (UC) rely on applications that are ingesting EGENIOUSS positioning outputs into their existing application environment. The UC show how the EGENIOUSS service is integrated into operational applications and which technical and organisational measures are needed to realise this integration. An analysis of application compatibility for each use case, together with data on user feedback and observed user behaviour collected during multiple test iterations with QField and Naviki, provides detailed insights into the practical application potential of EGENIOUSS. Test results and the methodology for user testing, encountered challenges and the lessons learnt from implementing and using EGENIOUSS within QField and Naviki are presented and discussed.

This contribution presents (1) a use-case-driven evaluation of smartphone-based positioning using the EGENIOUSS service in two representative urban scenarios: bicycle navigation and professional mobile surveying; (2) an assessment of the integration of multi-source positioning outputs into existing applications (Naviki and QField), including system compatibility, usability, and deployment considerations; and (3) an empirical analysis of positioning performance, user behaviour, and practical limitations observed during iterative field tests, highlighting challenges and lessons learnt when applying high-accuracy smartphone positioning in real-world urban environments

As autonomous vehicle (AV) technology advances from Level 2 assistance to fully self-driving operation, rising demands on positioning reliability and expectations for aviation-level safety highlight the need for certifiable navigation architecture suited to automotive use, especially in challenging environments such as urban canyons. In these settings, GNSS performance is degraded by non-line-of-sight (NLOS) conditions, poor satellite geometry, multipath, and interference, where blockage and propagation losses often weaken or eliminate satellite signals, which motivated the integration of additional sensors. Although AI-enabled navigation shows strong potential for multi-sensor fusion, its use in safety-critical functions is limited by certification complexity, limited explainability, and weak generalisation. Existing fusion methods using INS, VIO, LiDAR, and Radar involve trade-offs in drift, cost, memory, and robustness. Consequently, complementary sensors such as cameras and barometers are increasingly integrated into GNSS/IMU fusion frameworks to address GNSS outages, with barometric measurements specifically improving the weak performance in vertical direction under multipath and NLOS conditions for aerial applications. Automotive applications require similar enhancements in vertical performance across operational scenarios. Many state-of-the-art architectures use centralised fusion schemes, including filter-based approaches (e.g., EKF/UKF with adaptive tuning) and Factor Graph Optimisation (FGO), which struggle in complex, time-varying environments and create a single point of failure. These limitations support decentralised, federated architectures.

In this study, a novel approach is proposed that incorporates ESKF as local filters and FGO as a loosely coupled master filter within a federated fusion framework. To the best of the authors’ knowledge, this is the first attempt to combine ESKFs and FGO within a federated architecture to achieve improved positioning performance in both accuracy and robustness aspects. This framework leverages ESKF as local filters that integrate GNSS/IMU/Barometer and VO/IMU/Barometer, where linearisation of only the error states enhances stability and robustness in highly nonlinear subsystems. A master filter based on FGO then combines the local estimates, providing globally consistent smoothing, improved handling of nonlinear effects, and long-term drift correction through the incorporation of past states.
The proposed system is evaluated using the simulation scenarios generated using Spirent GSS7000 simulator integrated with AirSim with hardware-in-the-loop capability, able to generate multisensory navigation data, including vision data in realistic degradation scenarios. Testing datasets cover diverse complex scenarios, such as low-light and the effects of multipath in a dense urban setting. The results show that the FGO-based master filter delivers a 78% improvement in overall performance compared with a traditional ESKF-based solution. The integrated fusion framework can correct erroneous measurements across all sensing modalities and enhance the behaviour of the local filters under multiple faults, such as sensor noise, environmental dynamics, multipath, and corrupted sensor data. In the vertical axis, both filter configurations exhibit an 87% improvement when barometric measurements are incorporated into the architecture. Across different trajectory geometries, local filter 1 outperforms local filter 2, particularly in segments with sharp turns near buildings under GNSS outages. In these conditions, using FGO as the master filter yields a 55% performance gain over an ESKF master filter, demonstrating the efficiency and resilience of the proposed approach.

Modern prosperity depends on the availability of energy, universal connectivity, and mobility, delivered by synchronized communication networks, precise positioning, navigation and timing (PNT), stable power grids, and autonomous operations. These capabilities increasingly rely on space-based assets, which provide global coverage and precision. Their rapidly growing criticality is enabled by affordability, itself driven by dramatic reductions in launch costs and economies of scale. Beyond cost, a less explicit but foundational pillar is asset coherency: the spatiotemporal synchronization of constellation elements so that economically relevant functions act as a coherent system.

For example, despite dynamic Doppler-shifts and variable propagation delays between different satellites, the efficiency of satellite communication payloads heavily depends on tight frequency allocations, time multiplexing, coordinated frequency and beam hopping and Multiple-Input Multiple-Output (MIMO) techniques. Improved synchronization reduces guard bands and guard times, raising spectral efficiency (bits/s/Hz) and user capacity. Present operational targets are in the order of tens of nanoseconds for network synchronization. More stringent requirements - nanosecond today and trending to picosecond - apply to payloads generating navigation signals for ground and near-ground users, where timing stability directly determines PNT accuracy. Earth-observation payloads using interferometric synthetic aperture radar (InSAR) and radio-frequency geolocation demand microwave signal phase alignment within a fraction of the carrier wavelength.

Obviously, high-fidelity frequency and time transfer across large baselines is at the heart of any critical infrastructure payload for satellite constellations. This paper presents a versatile on-board timing and synchronization concept centered on a spaceborne Global Navigation Satellite System (GNSS) receiver as the core of a payload electronics platform compliant with the SpaceVPX standard.

The FoX NavRIX (“PinPoint”) receiver delivers high-accuracy real-time navigation by exploiting the GALILEO High Accuracy Service in combination with a Kalman filter, leveraging heritage from multiple flight missions to achieve high reliability and availability in New Space constellation environments. The receiver supports simultaneous multi-constellation, multi-frequency GNSS, features low acquisition/tracking thresholds, provides multiple antenna inputs with optional dislocated low-noise amplifiers, and includes self-calibration for exceptional long-term stability. Integrated with a multi-technology oscillator, PinPoint forms a robust on-board timing subsystem disciplined to GALILEO time by default, but resilient via acceptance of alternative radiofrequency or optical synchronization references, reverting to a local clock ensemble under loss of synchronization connectivity. For particularly critical applications, support for the GALILEO Public Regulated Service (PRS) is available. Additional FoX platform elements - advanced Software Defined Radio (SDR) and Single Board Computer (SBC) modules - enable interference cancellation and the implementation of wideband transmit and receive functions.

Through worked examples spanning low-Earth to cis lunar orbits, the paper quantitatively evaluates GNSS receiver-centered payload performance for communication, navigation, earth-observation and surveillance use cases. A few examples for a LEO-PNT payload are provided below (downladable from https://mft.beyondgravity.com/download/ef2ef264-493c-4f75-95f4-015940b4bb08).

In recent years, Global Navigation Satellite System (GNSS) spoofing incidents have occurred frequently in the real world, indicating that spoofing threats have become a non-negligible security concern for GNSS users. Small-delay spoofing attack degrades GNSS positioning accuracy by inducing minor deviations in satellite pseudo-range measurements, enabling it to evade most spoofing defense mechanisms and achieve successful intrusion. Evidently, small-delay spoofing poses greater information-security risks to GNSS users while simultaneously increasing the complexity and difficulty of spoofing defense.
Existing GNSS anti-spoofing methods encompass spoofing detection and spoofing mitigation. However, compared to the well-established detection techniques, spoofing mitigation research remains relatively limited. Moreover, existing mitigation approaches based on signal parameter estimation have been shown to be ineffective in suppressing small-delay spoofing attacks, necessitating additional countermeasures in the positioning solution process.
This paper proposes a novel small-delay spoofing mitigation framework that integrates brain-inspired Spiking Neural Networks (SNNs) with Factor Graph Optimization (FGO), eliminating the need for explicit spoofing detection. An end-to-end SNN-based GNSS measurement uncertainty modeling strategy is developed to estimate pseudo-range bias and observation noise uncertainties, using the positioning error obtained from the uncertainty-updated solution as the loss function to achieve label-free learning. The predicted uncertainties are subsequently incorporated into an FGO-based nonlinear estimator for factor-level measurement compensation and adaptive weight updating, effectively suppressing small-delay spoofing attacks and yielding high-precision estimates.
Experimental validation under controlled spoofing test scenarios confirms that the existing Multipath Estimating Delay Lock Loop (MEDLL)-based mitigation method fails to effectively suppress small-delay spoofing attacks, whereas the proposed SNN-enhanced Weighted Least Squares (SNN-WLS) method reduces the three-dimensional Root Mean Square Error (RMSE) of positioning solutions by 84.23%, and the proposed SNN-enhanced FGO method further decreases it by 18.51%. These results indicate that the proposed framework successfully alleviates spoofing-induced positioning errors and produces high-accuracy, smooth trajectory estimates. The proposed method provides real-time inference capability, energy-efficient edge deployment, and enhanced noise resilience in complex GNSS environments, making it particularly suitable for resource-constrained autonomous positioning applications facing small-delay spoofing threats.

Jamming and spoofing poses a significant problem for navigation systems. GNSS signals, which are crucial for navigation in vehicles, aircraft, and maritime operations, can be easily jammed, leading to navigation errors and potential accidents. As reliance on GNSS technology increases, the threat of jamming highlights the urgent need for resilient navigation systems and countermeasures to protect against such disruptions.
The Norwegian Mapping Authority (NMA) operates a national network RTK service, serving professional users who require high accuracy. NMA is also involved in european quality control projects, dealing with topics such as space weather, jamming and spoofing. To reach our goals of high quality and resilient GNSS positioning services in Norway, we cooperate closely with European partners.
Testing is critical to achieve more resilient navigation systems. To address this, the annual Jammertest event was established at Andøya in Norway by a group of Norwegian organizations. NMA is both part of the organizer group and a participant in the testing activites.
We will present our test activites at past Jammertests and also discuss our thoughts for coming events. Some of our results from Jammertest will be shown, covering both typical geodetic equipment and lower-cost receivers. Looking ahead, we discuss plans for future testing and hope to inspire discussions and feedback that can feed into the Jammertests to come.

Jamming and spoofing attacks on Global Navigation Satellite System (GNSS) receivers require dedicated mitigation strategies. The use of Controlled Reception Pattern Antenna (CRPA) is a well-established technique, that can be realized using multiple antenna elements of the receiver to form an antenna array to perform both beam steering to enhance high-elevation reception and null steering to suppress the presence of interferers. Recent advances have made antenna array-based GNSS receivers accessible to the civilian market in the form of Software-Defined Radio (SDR), particularly KrakenSDR, which offers five synchronized channels for use as an antenna array. However, the presence of multiple interferers may overcome such types of arrays, as they can only suppress as many interferers, as the number of antennas that make up the array. For applications in jamming stricken regions such as Eastern and Northern Europe, it is desirable to expand the capability and scale of such arrays to ensure continued GNSS reception for critical applications and do so at a reasonable cost.

To accomplish this, we present a unique injection signal to achieve synchronization between two units of KrakenSDR. The injection signal is transmitted from a separate SDR (injection-SDR), transmitted to a common channel of the two KrakenSDRs to align the time, frequency and phase of the incoming signal. With the reference channel common to both SDRs, this expansion effectively enables a scaling from five to nine antenna elements, with each additional KrakenSDRs adding four elements to the full array. This greatly limits the number of units involved relative to the total number of antenna elements, compared to conventional array upscaling.

The injection signal is transmitted via coax cable from a local unit, as opposed to an open-air transmission that is susceptible to jamming attacks. At the time of testing the setup, no external clock was used to generate injection signal as it was not available at the time. To alleviate this and enable flexible synchronization schemes independent of very accurate clocks, we introduced frequency modulated timestamps as part of the injection signal. Alongside chirp-signals that are well-defined in time and frequency to synchronize in these domains, the paired transmissions ensure synchronization at burst repetitions. The injection signal protocol is implemented in and operated from GNU Radio making it easy to use.

The initial results of operating the expanded array are promising, as the beam of the dual-array is clearly sharpened as compared to single-array beamforming for the controlled tests performed. We plan to include tests performed using the array constellation, tested under open-air GNSS jamming and spoofing at the annual Jammertest campaign in Andøya, Norway. In the planned paper, we hope to showcase the utility of the array, using it for localization as part of the driving tests performed at Jammertest 2025.

Supercorrelation®, a technology developed by Focal Point Positioning, has demonstrated excellent multipath suppression properties, significantly enhancing the navigation solution in challenging environments. So far the performance benefits have been shown for standard positioning applications, utilising common observables like code-phases and doppler [1] [2]. Focal Point Positioning has now extended this technology to high-precision applications. This extension focuses on generating improved accumulated delta range measurements, often known as carrier-phase.

In this work, we evaluate Supercorrelation® specifically for high-precision applications, emphasising the benefits to carrier-phase based positioning. We demonstrate how its multipath suppression capabilities extend to carrier-phase observables and, for the first time, present high-precision results achieved in challenging environments. The technology's advantages are shown in bridging and preventing cycle-slips, consequently enabling robust high-precision positioning in difficult settings. Commercially, this capability opens new avenues for Focal Point Positioning, particularly in autonomous vehicle navigation and advanced driver assistance systems (ADAS).

Our evaluation uses a combination of synthetic and real data. The testing is performed on an in-house software-defined receiver, initially developed as part of a European Space Agency (ESA) funded NAVISP project and subsequently extended independently to include high-precision functionality.


References:

[1] Garcia, J.G.; van der Merwe, J.R.; Esteves, P.; Jamal, D.; Benmendil, S.; Higgins, C.; Grey, R.; Coetzee, E.; Faragher, R. Development of a Custom GNSS Software Receiver Supporting Supercorrelation. Eng. Proc. 2023, 54, 9. https://doi.org/10.3390/ENC2023-15423

[2] Garcia, J.G.; van der Merwe, J.R.; Mwenegoha, H.; Esteves, P.; Benmendil, S.; Coetzee, E.; Ellis, J.; Eriksson-Martin, H.; Grey, R.; Higgins, C.; et al. Enhancing GNSS Robustness in Automotive Applications with Supercorrelation: Experimental Results in Urban Scenarios. Eng. Proc. 2025, 88, 75. https://doi.org/10.3390/engproc2025088075

The Galileo High Accuracy Service (HAS) became operational in January 2023, marking a significant advancement for users requiring high‑precision GNSS positioning. For registered users, it provides essential clock and orbit corrections, as well as code biases, covering both the Galileo and GPS satellite constellations. These valuable corrections are distributed through two convenient channels: the Signal‑In‑Space (SIS) interface using the Galileo E6 signal, and the Internet Data Distribution (IDD) method for broader accessibility. What makes this particularly powerful is its support for the Precise Point Positioning (PPP) technique, which allows users to achieve centimeter‑level accuracy in their positioning — even in remote areas without an internet connection, relying purely on the satellite signals they receive. According to the official HAS Service Definition Document, the system consistently delivers impressive positioning accuracy: better than 20 cm horizontally and below 40 cm vertically. In addition, Galileo HAS officially incorporates the Galileo Terrestrial Reference Frame (GTRF), ensuring that all user positions are precisely aligned within this global reference system.
Precise positioning relies on a unified reference frame such as ITRF2020, but GNSS constellations use their own frames, creating challenges for multi‑GNSS integration. This research analyzes the reference frame realization based on Galileo HAS corrections. These corrections, applied to the navigation messages, were used to generate orbits in the standard SP3 format.
The study utilized IDD data from the period after HAS began using the IGS20.atx calibration files (starting on May 15, 2025). These data were compared against final products provided in the ITRF2020 frame. For the dataset, seven transformation parameters were derived between the two frames. The obtained results not only demonstrate consistency with the ITRF frame but also allow for monitoring the realized frame’s stability over time through changes in the transformation parameters. The presented findings may help improve the consistency of final positioning results and ensure accurate transformations to national reference frames.

This study describes the characterization of signal performance degradation in the spaceborne GPS receiver and suggests mitigation techniques for signal attenuation intervals while the LEO satellite penetrates to jamming area.
Since the outbreak of the Russo-Ukrainian War on February 24, 2022, it was observed that the signal tracking performance of the satellite's GNSS receiver significantly degraded due to intensive GPS jamming within the affected regions.
Before the war, the receiver maintained a stable tracking performance with an average C/N0 of 41 dB-Hz.
However, after the war, the number of signal dropouts increased rapidly, resulting in a decrease in the average number of navigation satellites that could be tracked from 10.2 to 0.
On average, jamming-induced interference disrupted nominal GPS signal acquisition for approximately 18 minutes per pass, thereby disturbing DubaiSat-2’s mission operations.
To overcome these outages, an orbit propagation method could be considered to maintain a continuous navigation solution during signal loss, using either Two-Line Elements (TLE) or the most recent filtered navigation solution.
Given that the navigation receiver developed by Satrec Initiative for DubaiSat-2 incorporates real-time on-board orbit determination logic, the latter approach—leveraging high-precision real-time state vectors—was adopted.
Consequently, a functional enhancement was implemented, enabling the receiver to generate and provide an estimated orbit up to 20 minutes during signal anomalies.
This update ensures that the GPS receiver operability is maintained even during periods of non-nominal receiver operation.
This research presents a comparative analysis of operability before and after the war, along with the implementation of the proposed methodology.

Recent advances in satellite navigation systems reveal three major trends: the widespread adoption of multi-constellation Global Navigation Satellite System (GNSS), the continuous enhancement of single-constellation services through high-accuracy augmentation mechanisms such as High Accuracy Service (HAS), Satellite Based Augmentation System (SBAS), and ground-based reference networks, and the growing integration of non-GNSS satellite signals—particularly low Earth orbit (LEO) constellations—into positioning, navigation, and timing (PNT) solutions, exemplified by emerging systems such as European Space Agency (ESA). As a result, modern receivers are exposed to an unprecedented number of signals with increasingly heterogeneous quality characteristics.

While the availability of abundant high-quality signals improves positioning robustness, it also increases computational complexity and may degrade positioning performance if suboptimal signal combinations are used. Consequently, signal subset selection has become a critical component of next-generation PNT systems. This problem fundamentally involves two competing factors: the quality of individual measurements and the spatial distribution of selected satellites, such that reducing the number of signals tends to increase the risk of poor satellite geometry.

In this paper, we investigate the signal subset selection problem by explicitly incorporating GNSS signal characteristics related to elevation angle and carrier-to-noise density ratio (C/N₀). A comprehensive simulation framework is developed to generate multiple satellite visibility scenarios with varying constellation compositions, signal counts, and quality distributions. Based on these scenarios, we evaluate positioning performance across different subset selection strategies and propose an optimal selection algorithm that balances the number of satellites against geometric strength. Simulation results demonstrate that the proposed method effectively mitigates geometric degradation while leveraging high-quality measurements, resulting in improved positioning accuracy compared with conventional selection approaches.

Testing GNSS interference resilience can be a challenging task, particularly for larger systems that cannot easily be brought to suitable test environments such as Jammertest. We tested a selection of GNSS interference scenarios, mostly focused on timing, in a fully operating digital substation.

The substation uses two interconnected timeservers (station clock A and B), with station clock A acting as the primary clock and B as a backup. Both timeservers use GNSS as one of several sources of reference timing, including a Rb-oscillator for holdover capacity.

The test setup consisted of a software based GNSS simulator connected inline between the substation GNSS antennas and the substation clocks. The setup enables simulating true GNSS signals synchronized to the actual satellite signals received by the antenna, in combination with interference signals generated in the GNSS simulator. The setup allowed us to apply simulated signals to either one or both station clocks, which opens a wide range of possible test scenarios.

Clock ensembling is a key concept for future time-scale generation, enabling substantially improved robustness and stability compared with single master-clock architectures. It already underpins international reference time scales such as Universal Time Coordinated (UTC), which is computed from ensembles of atomic clocks operated in timing laboratories worldwide. By combining multiple clocks in an ensemble and appropriately weighting their contributions, the impact of failures or performance degradations of individual clocks on the generated time scale can be strongly reduced – but only if such faults are reliably detected and mitigated.
In the Robust Precise Timing Facility (RPTF) of the German Aerospace Center’s Galileo Competence Center (DLR GK), we implement a Composite Clock Algorithm (CCA) that operates on a mixed ensemble of four active hydrogen masers and four cesium clocks. The clocks are continuously measured against each other on redundant measurement chains, which increases availability of the time scale and enables consistency checks within the ensemble. Such heterogeneous ensembles not only increase robustness, but also exploit the complementary stability characteristics of the participating clocks over different averaging times. As a result, the generated ensemble time scale outperforms every individual clock in the ensemble.
In our implementation, we follow the Kalman-filter-based approach introduced by Brown [1] and later refined by Greenhall [2] to generate a “paper” time scale known as the Implicit Ensemble Mean (IEM). However, failures or anomalies in individual clocks directly corrupt the estimates feeding the IEM and, if left unaddressed, can significantly degrade the stability and robustness of the composite time scale.
The proposed CCA employs two Kalman filters. The first estimates the deviation of each individual clock from the IEM. The second computes steering commands for an active hydrogen maser and an offset generator to realize a physical approximation of the IEM as the output time scale. To ensure the superior stability of this composite time, we developed a comprehensive fault-detection, identification and isolation (FDII) framework. It comprises (i) detection of the presence of a fault, (ii) identification of the faulty clock within a certain timeframe, and (iii) fault-handling strategies such as temporarily removing affected clocks from the ensemble. While the literature describes FDII concepts based, for example, on statistical tests of Kalman-filter residuals, practical operation has shown that these approaches require substantial adaptation to real hardware behavior and measurement conditions. We therefore designed additional anomaly-detection strategies tailored to our ensemble and validated them in both simulations and long-term tests in the RPTF with real clocks under operational conditions. The presentation will focus on these FDII strategies and share experiences and results from the long-term experiments in the RPTF.

[1] K. R. Brown, “The theory of the GPS Composite Clock”, Proceedings of the 4th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GPS 1991), pp. 223 – 242, Institute of Navigation, 1991
[2] C. A. Greenhall, “A Kalman filter clock ensemble algorithm that admits measurement noise,” Metrologia, vol. 43, no. 4, pp. 311–321, 2006

Modern aircraft and unmanned aerial vehicle (UAV) navigation systems rely heavily on physical control interfaces, remote controllers, and predefined autonomy pipelines. While these systems have achieved high levels of reliability, they often impose operational constraints in dynamic or time-critical scenarios, such as search and rescue, human–machine teaming, and mixed-autonomy environments. In particular, conventional controller-based navigation introduces limitations related to operator workload, reaction time, and interface dependency, which may reduce system flexibility and situational adaptability. This challenge becomes more pronounced as aerial platforms increasingly operate in complex environments requiring intuitive, rapid, and adaptive human input.

A critical gap in current navigation paradigms is the lack of natural, vision-based human interaction mechanisms that can complement or partially replace traditional control hardware without compromising safety or responsiveness. Existing approaches primarily focus on autonomy or sensor-based navigation, while comparatively little attention has been given to leveraging human intent captured through visual perception as a direct navigation input modality.

In this paper, we propose a computer-vision–driven navigation interface that enhances UAV control by translating human hand gestures into real-time navigation commands. The proposed framework integrates deep-learning–based pose estimation with control-system design to enable hands-free and controller-independent drone navigation. Using a monocular camera, the system extracts finger and hand keypoints via a PoseNet-based model, classifies predefined gestures, and maps them to aircraft-level motion commands such as directional movement, hovering, ascent, descent, and landing.

The system architecture is composed of three core components: visual perception and gesture recognition, command generation and validation, and UAV communication and execution. The model is implemented using Python and OpenCV, with training and optimization performed using TensorFlow-based tools. Special attention is given to low-latency processing to ensure responsiveness suitable for real-time aerial navigation.

Experimental evaluations demonstrate that the proposed method enables stable and intuitive UAV navigation while significantly reducing reliance on traditional control devices. The vision-based interface shows robustness under varying environmental conditions and maintains consistent command interpretation during flight.

This work contributes a novel human-in-the-loop navigation paradigm that augments existing aircraft and UAV control systems through computer vision. By reducing dependency on physical controllers and enabling intuitive visual interaction, the proposed approach offers a promising direction for next-generation aerial navigation systems, with potential extensions to collaborative UAV operations and future human–autonomy integration in aerospace applications.

The vulnerabilities of Global Navigation Satellite Systems (GNSS) have spurred the development of Low Earth Orbit (LEO) positioning, navigation, and timing (PNT) systems. LEO satellites offer signals with power levels 25–35 dB higher than GNSS and high dynamic geometries that accelerate the observability of state parameters. Among these signals of opportunity, the Orbcomm constellation is a candidate due to its open data information and continuous signal transmission, from which the carrier phase measurement can be extracted. However, transitioning from Doppler-based positioning to precise carrier phase navigation presents significant theoretical challenges.
The primary objective of this study is to design, implement, and evaluate a signal processing architecture capable of acquiring and tracking Orbcomm downlink signals to enable precise positioning. This involves developing an acquisition algorithm that leverages the properties of symmetric differential phase shift keying (SDPSK) modulation to generate multi-frequency carrier phase measurements using a zero-IF architecture. Furthermore, this study compares the tracking robustness of a third-order phase-locked loop (PLL) against a linear Kalman filter (LKF) loop, serving the development of a precise carrier phase-based positioning algorithm.
Utilizing the characteristic of the SDPSK of the signal, a deterministic code sequence can be generated. Then, an FFT-based parallel code phase search is combined with a constant false alarm rate detector for acquisition. For tracking, two architectures are implemented: a conventional third-order PLL assisted by a frequency-locked loop to handle high Doppler rates, and an LKF tracking loop designed to optimally estimate signal carrier phase, frequency, and frequency rate. Both methods employ a zero-IF architecture, where the accumulated carrier phase is generated by numerically integrating the carrier numerically controlled oscillator output, ensuring cycle continuity. For positioning, this study employs carrier phase navigation algorithms that utilize the generated zero-IF measurements. The positioning framework is primarily constructed using Least squares and extended Kalman filter estimators to solve for the receiver's state. Furthermore, the research investigates potential strategies for carrier phase ambiguity resolution that are suitable for the Orbcomm signal structure, aiming to refine positioning precision further.
To quantitatively evaluate the proposed signal processing architecture, the tracking integrity is primarily assessed using the Phase Lock Indicator (PLI) and the Carrier-to-Noise ratio, followed by an analysis of the phase noise characteristics. It is anticipated that the LKF tracking loop will demonstrate better performance in maintaining a stable PLI, particularly under conditions where the conventional third-order PLL may experience lock loss or significant signal degradation. Additionally, the LKF is expected to yield reduced phase jitter, thereby enhancing measurement quality. This research establishes a fundamental framework for carrier phase positioning using Orbcomm LEO satellites. The implementation of the LKF tracking loop validates the ability to extract carrier phase observables from low-symbol-rate SDPSK signals. These results confirm the feasibility of extracting carrier phase measurement and performing positioning from Orbcomm satellites.

GNSS positioning accuracy in challenged environments, such as urban canyons or foliage-obstructed areas, relies heavily on the fidelity of the stochastic model and the algorithm's capability to mitigate local signal distortions like multipath and Non-Line-Of-Sight (NLOS) reception. This paper presents a comprehensive analysis and a direct comparison of different observation weighting strategies implemented within an Extended Kalman Filter (EKF) navigation algorithm.
The filter is designed using a Position-Velocity (PV) model with an 8-state vector, estimating the receiver’s 3D position, 3D velocity, clock bias, and clock drift, while processing raw code pseudorange observables. The study focuses on the optimal construction of the measurement noise covariance matrix (R). Initially, the a priori residual variance (σ^2) is modeled by comparing three base weighting techniques: (i) a model dependent on satellite elevation; (ii) a model dependent on the measured Carrier-to-Noise density ratio (C/N0); (iii) a linear combination of both metrics to capture different aspects of signal quality.
Subsequently, the stochastic model is dynamically refined to address specific error sources: the Time-Differenced Code-Minus-Carrier (TD-CMC) observable is employed for the identification and mitigation of multipath effects, while the IGG-III reweighting method is applied for the robust suppression of Non-Line-Of-Sight (NLOS) signals.
Experimental validation was conducted by processing RINEX files acquired via a Piksi Multi GNSS Module in both static and dynamic scenarios. Results demonstrate that the combined approach, assisted by IGG-III logic, yields superior performance in minimizing estimation error compared to the other considered methods.

Doppler-only positioning using Low Earth Orbit (LEO) satellites has emerged as a promising alternative to traditional Global Navigation Satellite Systems (GNSS), offering resilience through signal diversity and favorable geometric properties. Whereas the theoretical framework for Doppler-based navigation has been established, efficient selection of a satellite subset providing the best positioning solution remains an open problem. Unlike pseudorange-based GNSS, where spatial volume maximization provides effective satellite selection, recent empirical studies have shown that this approach performs no better than random selection for Doppler Geometric Dilution of Precision (D-GDOP). This work addresses this gap by analyzing the geometric structure of the Doppler observation Jacobian to characterize optimal satellite configurations and provide theoretical foundations for subset selection algorithms.

The standard 8-state Doppler navigation formulation estimates position, velocity, clock bias, and clock drift. The observation Jacobian includes a clock bias sensitivity column arising from the coupling between timing errors and the changing satellite geometry. We analyze the structure of this coupling term to understand its correlation with position and velocity columns in the Jacobian. Additionally, we introduce a reduced 7-state formulation that excludes clock bias estimation, serving as an analytical tool to isolate geometric sensitivities from the complications introduced by timing estimation. This simplified formulation may facilitate the development of satellite selection algorithms by providing insight into the underlying geometric structure.
We formulate the GDOP minimization problem incorporating orbital mechanics constraints, including Clairaut's relation for orbital heading, elevation-dependent line-of-sight rates, and satellite reachability conditions. The optimization framework considers both fixed-altitude and variable-altitude scenarios to investigate how altitude diversity affects the Jacobian correlation structure. Numerical simulations validate the theoretical predictions and reveal the characteristics of optimal satellite configurations.

Analysis of the Jacobian structure shows that the clock bias coupling term can exhibit correlation with multiple state columns depending on the satellite geometry, which helps explain why traditional volume-based selection metrics fail for D-GDOP minimization. The variable-altitude optimization reveals that altitude diversity provides a mechanism for improving the conditioning of the 8-state estimation problem, while the 7-state formulation exhibits optimal geometries similar to the full 8-state system, justifying its value in geometric analysis.

These results provide geometric insight into what constitutes favorable satellite configurations for Doppler-based navigation. The identified optimal patterns ultimately offer criteria for developing computationally-efficient satellite selection algorithms or augmentation systems.

In aeronautical applications ensuring highly reliable navigation is essential to flight safety. While Receiver Autonomous Integrity Monitoring (RAIM) is a standard approach to ensure operational integrity for GNSS (Global Navigation Satellite Systems), reliance on RAIM alone can lead to ‘RAIM holes’ – areas where sufficient integrity cannot be guaranteed at all times due to GNSS constellation limitations. Another threat to navigation reliability is malicious interference with the satellite signals in the form of jamming or spoofing. If GPS (Global Positioning System) is the only GNSS used, as is often the case in aviation, this is particularly critical.
A possible solution is an advanced approach like ARAIM (Advanced Receiver Autonomous Integrity Monitoring) to incorporate other constellations or ground stations. ARAIM concepts are currently under research but not yet standard in aeronautical applications. Another way to increase integrity metrics are integrated systems that use secondary sensors for cross-validation of GNSS signals. Typically used sensors for this application are IMUs (inertial measurement units).
This presentation explores how incorporating information from a Doppler radar as an additional aiding sensor can extend the integrity monitoring capabilities of a tightly coupled Inertial/GPS navigation system utilizing a solution separation approach to detect faulty satellites. The sensor model is implemented as four symmetrically aligned relative distance measurements in the respective line-of-sights like a Doppler-Radar would approximately measure in its antenna boresight directions. Failure modes that are examined are ramp errors on the horizontally hardest to detect GPS-satellite in adverse geometries as well as failures of the Doppler sensor itself.
The presentation begins with a brief overview of current state-of-the-art integrity monitoring in tightly-coupled Inertial/GNSS systems. Subsequently, we demonstrate how the sensor data is modelled and integrated in the system. Finally, we present simulation results comparing the performance of the integrated system with Doppler information to a receiver-only solution and a standard Inertial/GNSS configuration, highlighting the potential for enhanced navigation performance.

The Algerian Satellite-Based Augmentation System (AL-SBAS), developed by the Algerian Space Agency (ASAL), broadcasts SBAS messages through the national geostationary satellite Alcomsat-1 (24.8°W) in full compliance with International Civil Aviation Organization (ICAO) standards. Unlike wide-area augmentation systems such as EGNOS, AL-SBAS is designed as a regional system dedicated to supporting GNSS-based navigation across Algeria. Correction messages are generated by the Mission Control Center using GPS measurements from a national network of Reference Stations distributed across the country.
This study presents an initial performance assessment of AL-SBAS for civil aviation by comparing APV-I availability with EGNOS. APV-I availability is calculated as the percentage of epochs where the Horizontal and Vertical Protection Levels (HPL < 40 m, VPL < 50 m) remain below ICAO alert limits.
Results indicate that AL-SBAS APV-I availability was below 96 % in January 2023, whereas EGNOS exceeded 99 % for the same period. Although AL-SBAS is still under development, these results are encouraging and demonstrate the system’s potential to support future aviation operations within Algeria.

Satellite positioning has become a fundamental element of modern life, supporting from everyday navigation to highly safety-critical applications. For many years, Global Navigation Satellite Systems (GNSS) like GPS and Galileo have been the primary providers of Positioning, Navigation, and Timing (PNT) services. However, because these systems rely on relatively weak signals transmitted from distant satellites, they are increasingly vulnerable to various types of disturbance. These limitations have shifted attention toward satellites operating in low Earth orbit (LEO), which are rapidly emerging as a more resilient solution expected to offer stronger signals. LEO-based PNT is currently the subject of active discussion, as these systems are still under development. It remains an open question whether they will achieve supremacy over traditional GNSS or primarily complement existing navigation systems.
In this research, we address the key issue of radio frequency interference (RFI), specifically jamming, which has not yet been addressed in the context of LEO, and we evaluate LEO performance compared to traditional GNSS. Our approach relies on simulating both GNSS and LEO constellations using Skydel simulation software and generating the corresponding GPS (L1 C/A), Galileo (E1) and LEO (E1) signals. Interference is then added to these clean generated signals. We selected multiple types of jamming signals with varying characteristics that represent typical interference affecting GNSS, namely single-tone signals, frequency-modulated signals, frequency-hopping signals and chirp signals, to investigate how these different interference signals affect system behavior.
We have promising preliminary results showing that LEO signals are more resilient to chirp interference than GNSS constellations. In the final paper, we will extend this stidy to other interference scenarios that are still under analysis. For each case, the resulting signals will be processed using an open-source software-defined receiver (SDR) to evaluate the impact at different stages of the signal-processing chain: directly on the raw IQ samples using the Running Digital Sum (RDS), and after correlation using C/N₀, as well as the outputs of the Phase-Locked Loop (PLL) and Delay-Locked Loop (DLL).

Batch processing methods have bring important performance benefits to terrestrial navigation. For example, Kalman Filtering (KF) can improve the navigation performance of a terrestrial ranging system, to the point where it can enable new types of operations, compared to the raw performance of a snapshot equivalent. This is particularly important for systems that fuse inputs from multiple sensors.
An open challenge for batch processing has typically been the dependency of navigation performance on the trajectory itself. The dependency of position estimates on the history of the true position is fundamental and no linearizations or separation of variables can simplify the problem.
In planning terrestrial navigation infrastructure, such as a network of Distance Measuring Equipment (DME) ground stations. For planning purposes it is helpful to understand the achievable navigation performance at any given point in space. But any of those points in space can be reached via an arbitrary trajectory, with many degrees of freedom, but the trajectory is critical to the prediction of the achievable navigation performance at that point.
This paper will propose a solution to the conundrum by modeling local performance as invariant. The method enables better planning of the infrastructure, as the impact of changes to the terrestrial infrastructure reflect immediately in any map of achievable performance.
The method is a new tool in the design of navigation ground infrastructure. The paper will focus on conventional navigation, using DME to reach various Area Navigation (RNAV) and Required Navigation Performance (RNP) service types. Strictly speaking, however, the method generalizes to other environments and systems.
For the application of non-precision approaches (NPA) with either RNAV or RNP, the preliminary results show that the horizontal error of a DME-based navigation solution can be reduced. In the initial cases analyzed for this paper the reduction has gone from several hundred meteres horizontal error to 50 m or better.
This kind of performance improvement can make the difference between providing an en-route service like RNAV 1, without batch processing, or an NPA service like RNP 0.3. For future applications it is also important to point out the value of this methodology in planning an alternate capability for autonomous aircraft, be they Advanced (Personal) Air Mobility (AAM) vehicles for human transportation, or autonomous delivery drones. Both types of Unmanned Aerial Vehicles (UAV) will require terrestrial navigation to complement satellite navigation.

Precise Point Positioning (PPP) enables centimeter- to decimeter-level positioning using a single Global Navigation Satellite System (GNSS) receiver and precise orbit and clock products, but its widespread adoption has long been constrained by slow convergence, often requiring tens of minutes to achieve full accuracy. Since 2020, augmenting PPP with signals from Low Earth Orbit (LEO) satellites has emerged as a promising solution to this limitation. Owing to their low altitudes, rapid apparent motion, stronger signal power, and improved geometry, LEO satellites fundamentally enhance parameter observability and accelerate ambiguity convergence. This review provides a comprehensive synthesis of research progress on LEO-augmented PPP from 2020 to 2025, covering simulation-based performance analyses, constellation design studies, and the first real-world experimental demonstrations using in-orbit LEO navigation satellites. Simulation studies consistently predict order-of-magnitude reductions in PPP convergence time, while recent experimental results with prototype LEO systems confirm convergence time reductions exceeding 50–70% with as few as one to two LEO satellites, alongside modest improvements in positioning accuracy. The review further examines the role of LEO augmentation in PPP ambiguity resolution and PPP-RTK frameworks, highlighting synergistic effects that enable near-instantaneous centimeter-level positioning under favorable conditions. Key technical challenges, including real-time LEO orbit and clock determination, hardware bias calibration, signal design, and system integration—are critically assessed in the specific context of their impact on PPP performance. The review further discusses future research directions focused on improving the effectiveness and reliability of LEO-augmented PPP, such as enhanced orbit and clock modeling, bias mitigation strategies, optimized constellation configurations for faster convergence, and validation in challenging environments including urban and high-latitude regions. Overall, the reviewed studies demonstrate that LEO augmentation has a substantial and quantifiable impact on PPP convergence behavior and robustness, offering a viable pathway to significantly reduce initialization time and improve solution availability for high-precision GNSS positioning.

Navigation software typically deals with a single sensor or aims at sensor fusion by combining two or more complementary sensors. However, few software solutions are flexible enough to support multiple configurations or let one change between different hardware setups without rewriting the code.
Flow-based programming (FBP), which is the underlying concept of our software INSTINCT (Topp et al., 2025), allows for a much higher flexibility than existing software packages regarding the combination of different sensors and adaptability to user-defined requirements. GNU Radio, IoT, or robotics are just a few examples that have already exploited the potential of FBP. In addition to GNSS receivers and Inertial Measurement Units (IMUs), INSTINCT is capable of processing barometers and perform WiFi ranging, using the fine time measurement introduced in the IEEE standard 802.11mc. All of these sensors can be processed individually, or fused together to obtain measurements. The sensors can be interfaced easily using INSTINCT's graphical user interface, which removes the need to write specific code.Inertial sensors can be processed individually, or IMU arrays can be combined, leveraging the effects of noise reduction and improving robustness against sensor outages or failures. Results can be displayed in INSTINCT's own plotting node, output in human-readable files, or streamed as UDP packages for external use. In addition, INSTINCT features a wide range of nodes that allow signal processing, such as filtering or decimation, manipulation, or forming combinations of different signals or results, which is especially useful when comparing different navigation systems or configurations. INSTINCT follows the flow-based paradigms, which implies that parallelism can be achieved easily. This turns out to be very useful when implementing solutions that rely on filter-bank schemes or have several filters running in parallel for different purposes. Since observations can also be written to standardized formats, like for example the Receiver Independent Exchange Format (RINEX), INSTINCT can also be used as a converter or data-logger, both in real-time and post-processing. The software is open-source, runs on Linux, macOS, and Windows, and can even be used on small single-board computers, like a Raspberry Pi, which is especially useful for autonomous flying or driving applications. INSTINCT is easily extendable with new sensors since this only requires the implementation of vendor-specific interfaces, while the rest of the processing chain can be used from well-proven flows that are continuously verified through unit-tests. Our presentation will cover an overview on the idea of FBP, the main features of INSTINCT, and examples on how the software can be used to easily create navigation solutions without the need to write a single line of code.

Smartphone GNSS measurements are of lower quality than those from commercial GNSS receivers due to challenging reception conditions and hardware constraints. To encourage research on accurate navigation from such low-quality data, the Google Smartphone Decimeter Challenge (GSDC) was organized, in which a Factor Graph Optimization (FGO)-based method [1] achieved substantially better positioning accuracy than conventional techniques. FGO represents the relationship between state variables and observations as a factor graph and computes navigation solutions through iterative optimization. However, its high computational cost makes real-time smartphone-based navigation challenging under limited computational resources, motivating environment-specific factor configurations that balance positioning accuracy and computational load.
This study aims to derive environment-specific optimal factor configurations for Samsung smartphones through a systematic analysis of accuracy-computation trade-offs. We first quantify measurement quality and validate our methodology using Pixel 5 GSDC data, then apply it to characterize measurements of Samsung chipsets. We classify the positioning environment as urban, suburban, or open-sky based on user speed, the number of visible satellites, and signal strength. Starting from a baseline FGO composed of the Code and Doppler factors, we add the Motion factor and the Clock-relation factor, among other factors, and analyze the trade-off between positioning accuracy and computational cost. As a result, the Motion factor improves accuracy by approximately 2.7% in urban environments at the expense of a 107% increase in computation, whereas in open-sky environments, it yields about a 1.6% accuracy improvement with a 32% increase in computation. The Clock-relation factor provides about a 3.6% accuracy gain with only a 5% increase in computation in urban environments, and about a 1.0% accuracy gain with a 5% increase in open-sky environments. These results show that accuracy improvement and computational cost do not scale linearly and that the efficiency of each factor depends on the environment.
Building on this methodology, we quantify the measurement quality of the Samsung GNSS chipset to design an FGO framework optimized for Samsung smartphones. Using single- and double-differencing techniques with a commercial-grade receiver, we obtain, under open-sky environments, a code multipath error standard deviation of 7.10 m, Doppler noise of 8.42 cm/s, and carrier-phase noise of 12.37 mm. Applying the environment-classification methodology established from the Pixel 5 analysis, we classify Samsung measurements into urban, suburban, and open-sky environments. Finally, by integrating the environment-dependent measurement-quality characteristics of the Samsung smartphone with the factor-level accuracy–computation analysis derived from the Pixel 5 data, we propose an environment-adaptive factor selection strategy and a real-time FGO navigation framework tailored to Samsung smartphones.

[1] Suzuki, T. (2024, September). Second place winner of the smartphone decimeter challenge: An open-source factor graph optimization package for gnss and imu integration in smartphones. In Proceedings of the 37th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2024) (pp. 2703-2713).

Chip-Scale Atomic Clocks (CSACs) play a crucial role in high-precision timing systems due to their compact size and exceptional short-term stability. However, maintaining accurate timing capabilities when CSACs operate autonomously in GNSS-denied conditions remains a significant challenge. Specially, under scenarios such as GNSS signal blockage or jamming, the free-running CSAC bias accumulates rapidly due to inherent frequency aging and environmental disturbances. Existing prediction methods, including traditional stochastic models, are inadequate in capturing the highly complex, non-linear, and non-stationary characteristics of free-running CSAC time series. Therefore, this paper proposes an enhanced Wavelet Neural Network (WNN) for high-accuracy, multi-step CSAC bias prediction. By incorporating the multi-scale analysis capabilities of Wavelet Transform, the WNN can effectively capture both slow, long-term drift and rapid, transient frequency fluctuations in the clock data. This enables robust modeling of the intricate characteristics of CSAC behavior in GNSS-denied conditions. On real-world undisciplined CSAC data, the proposed method achieves a prediction Root Mean Square Error (RMSE) of less than 10 nanoseconds. Compared with traditional models, the enhanced WNN significantly reduces error divergence and extends the high-precision holdover time. This paper provides an effective method for enhancing the resilience of CSAC timing systems in challenging environments.

This research loosely integrates vector tracking with a dynamic model (DM) to enable reduced-kinematic orbit determination for low Earth orbit (LEO) nanosatellites. The operational scenario of spaceborne global navigation satellite system (GNSS) receivers is characterized by severe Doppler shift, varying carrier-to-noise density ratio (C/N₀), and constantly fluctuating GNSS satellite visibility. Compared to the conventional scalar tracking loop (STL), the VTL technique has demonstrated superior performance in signal tracking and navigation processing in low signal-to-noise ratio (SNR) environments and high-dynamic scenarios. This makes VTL well-suited to address the challenges encountered by spaceborne GNSS receivers on nanosatellites operating in LEO. Given the growing scale of LEO mega-constellations formed by small or nanosatellites, there is significant motivation to develop a VTL-based GNSS receiver algorithm optimized for the spaceborne scenario.

Despite its benefits, VTL has the drawback that a single faulty channel can adversely affect all other healthy channels due to the nature of its information fusion. Therefore, this research aims to address this inherent issue by integrating VTL with a DM and exploiting their respective advantages. The short-term accuracy of the DM can rectify occasional false estimations in VTL caused by erroneous measurements, while VTL can restrain the long-term drift of the DM. The loose coupling approach ensures that if the solution from one estimator becomes unavailable or degraded, valid results can still be obtained from the other estimator without cross-contamination.

Unlike the widely used reduced-dynamic technique for orbit determination, in which the DM is primarily responsible for orbit propagation and GNSS observations serve as auxiliary corrections to long-term drifts, this research places the GNSS-based VTL in the dominant role with the DM providing assistance. Hence, the algorithm is named dynamic model-assisted vector tracking (DMA-VT). The rationale for this approach is that for a wide range of applications, such as LEO-based positioning, navigation, and timing (PNT), the CubeSat's timing accuracy is equally as critical as position and velocity for broadcasting. VTL can be configured with a higher update rate compared to the DM to maintain fresh estimates of the onboard clock state.

Preliminary results demonstrate that DMA-VT can successfully track simulated spaceborne GNSS data with greater stability compared to STL, indicating robust tracking aided by filter-based model propagation and adequate dynamics prediction from navigation state feedback. Both position and velocity estimations show noticeable and consistent improvements compared to STL, with mean position and velocity root-mean-square errors (RMSEs) of 3.16 m and 0.29 m/s, respectively.

Kalman filtering (KF) is one of the reference state estimation algorithms in navigation applications. The navigation filter is highly sensitive to the tuning of the filter's noise covariance matrices which describe a physical phenomenon that should be accurately represented. Adjusting the parameters of the process noise covariance is among the most challenging aspects when implementing a KF. This matrix is typically tuned empirically based on prior knowledge of the system's overall dynamics and is not often constant over time. In this case the KF cannot consider sudden changes in the dynamics of the system when it involves strong jerk. However, adapting the filter to the vehicle dynamics requires to consider a time-varying, non diagonal process noise covariance matrix. To tackle this problem, a class of estimation techniques provide a wide range of methods estimating the state and the filter noise parameters, jointly or independently. However, they face challenges to capture correlated noises and time varying covariance matrices. To attempt to overcome these limits, a more recent approach proposes to leverage AI-aided Kalman Filtering techniques, hybridised with Recurrent Neural Network (RNN). These architectures are built to sustain incomplete stochastic models, learning the covariance matrices from the filter observations. By nature, the state-of-the-art version of RNN is formulated to predict varying covariance matrices, either diagonal or reconstructed complete matrices. Yet, the Symmetric Positive Definite (SPD) property of the covariance matrix is not verified.
We propose a benchmark of two types of RNN architectures predicting covariance matrices with correlation terms. Such assumption is crucial to capture the imprecision on the filter process model. We thus design a RNN based on modified Cholesky decomposition, capturing the SPD property of the predicted covariance matrices. It is compared to a conventional RNN where covariance matrices are learned as vectors. We benchmark their inference performances for prediction tasks. To perform this evaluation we build a new benchmark along with its semi-synthetic dataset. Real GNSS satellites positions and error terms are combined with simulated trajectories to reconstruct the GNSS observables. This allows to create 3 different scenarios to experiment diverse sequential behaviour ranging from stationary vehicle dynamics to non-stationary process noise covariance matrix.

This study proposes the configuration and design methodology of an Attitude Determination and Control System (ADCS) as a key component for ensuring the reliability and navigation accuracy of a Korean low Earth orbit (LEO) satellite navigation system. LEO satellites offer navigation advantages such as stronger signal power, pronounced Doppler variations, and improved satellite visibility due to their high orbital velocity and proximity to Earth. However, these benefits also introduce challenges, including highly dynamic operating conditions and limited platform resources, which necessitate precise attitude stabilization. In this work, an integrated ADCS architecture is developed by combining an attitude determination algorithm based on gyroscopes and star trackers with an attitude control structure employing reaction wheels and magnetic torquers. In addition, antenna-pointing requirements for maintaining navigation signal directionality are analyzed, and a closed-loop controller is designed and validated to satisfy these constraints. The results of this study provide a foundational lightweight and low-power ADCS architecture suitable for deployment in future large-scale Korean LEO navigation satellite constellations.

The rapid growth of low Earth orbit (LEO) satellite networks creates new demands for user terminals. Signals from fast-moving LEO satellites experience extreme and rapidly changing Doppler shifts, often exceeding ±100 kHz with change rates up to 1 kHz/s. Standard receiver tracking loops, which use narrow bandwidths (e.g., 15 Hz) to filter noise, are too slow to follow these dynamics. This leads to large carrier phase errors and frequent loss of signal lock.
To help with the tracking loop, receivers can use satellite orbit data (ephemeris) and motion sensor (Inertial Measuring Unit (IMU)) information to predict the main Doppler shift. However, this prediction is not perfect. Errors from the IMU, small inaccuracies in orbit data, clock imperfections, and noise create an unpredictable residual error in the prediction. This residual error still stresses the tracking loop and limits performance.
This paper proposes a novel method to address this problem by employing a Long Short-Term Memory (LSTM) neural network to learn and forecast the residual error of Doppler prediction. A large simulated dataset with rich feature information including multiple challenging conditions is constructed to support model training. The LSTM network learns from historical sequences of aided Doppler predictions, IMU measurements, and receiver-related states (e.g., signal-to-noise ratio, autocorrelation function (ACF) shape, loop outputs), as well as satellite-related parameters such as elevation and azimuth angles. By capturing the complex and nonlinear temporal dependencies that conventional models fail to represent, the LSTM effectively predicts the future Doppler residual. The residual is then used to refine the Doppler prediction of fundamental physical model, thereby enhancing the overall accuracy of the tracking loop.
The dataset is divided into training and testing subsets, where the LSTM model is trained using the training set and evaluated on the testing set. Experimental results demonstrate that the proposed LSTM-augmented Doppler residual prediction method substantially reduces the Doppler frequency tracking error. Compared with a conventional aided receiver, the proposed method significantly decreases the likelihood of tracking failures, particularly during rapid satellite passes. Moreover, thanks to its data-driven modeling of dynamics, channel variability, and loop state, the system exhibits enhanced robustness under challenging conditions, including weak signals, partial blockage, and diverse velocity profiles. These improvements enable more reliable and high precision signal tracking, which is essential for advanced LEO satellite-based positioning, navigation, and communication applications.
The innovations of this work lie in three main aspects. First, a machine learning-based Doppler residual prediction framework is introduced to enhance LEO Doppler compensation, enabling data-driven correction of prediction errors and improving high dynamics signal tracking performance. Second, the approach integrates satellite dynamics, channel characteristics, and tracking loop states as multi-domain features, providing a comprehensive representation of the tracking environment. Third, the proposed method exhibits enhanced robustness under challenging conditions, including weak signal environments, signal blockage, and varying velocity scenarios.

Monocular Visual Odometry (VO) estimates an agent's relative pose (position and orientation) from a sequence of images captured by a monocular camera. It is widely studied in robotics and used as a component of multi-sensor and autonomous navigation stacks, although pure monocular systems are less common in deployed autonomous vehicles because they suffer from scale ambiguity and drift. Most VO estimation has relied on geometric constraints derived from image sequences, a methodology that requires significant engineering effort and careful fine-tuning of modules, including feature extraction, matching, motion estimation and refinement. Recently, deep learning–based methods have emerged as a viable alternative, particularly when supported by large datasets and supervised training with metric ground truth or when combined with auxiliary sensors. Among these, techniques based on vision transformers have shown promise for estimating the 6-degrees-of-freedom (6-DoF) pose of an agent. While these methods can model global context via self-attention, their generalization depends strongly on training data and supervision; they may still suffer from long-term trajectory drift, sensitivity to dynamic objects and challenges in exploiting subtle frame-to-frame cues over extended trajectories. There also exist flow-based VO methods such as classical sparse Lucas–Kanade VO, dense optical-flow odometry, and recently proposed learned methods like LiteFlowNet. Dense optical flow provides per-pixel motion cues that can be informative for motion estimation, but flow estimation is itself challenged by low texture, occlusions and dynamic scenes; flow-based VO methods can still accumulate drift over long trajectories without geometric or metric constraints.
We propose FlowVoT, a GMFlow-based visual odometry framework with a Transformer-based motion estimator. The core novelty of FlowVoT is its use of GMFlow, a global matching-based optical flow network, as a pretrained dense motion input to a lightweight spatio-temporal Transformer. Unlike prior flow-based VO methods such as DF-VO, FlowVoT does not rely on explicit depth estimation or inertial sensors. It also differs from Transformer-based VO models like VoT and ViTVO in that it does not require a pre-trained visual encoder or depth input, but instead uses the pretrained GMFlow output as direct input to the Transformer, reducing complexity and computational burden
FlowVoT is an end-to-end trainable framework that integrates pretrained GMFlow optical flow with a lightweight Transformer to predict 6-DoF relative poses. Our approach leverages the strengths of dense per-pixel motion cues for challenging conditions (e.g., fast motion and occlusions), while employing spatio-temporal attention to capture global representations and improve motion estimation. We evaluated the proposed method on the widely used KITTI visual odometry dataset and report quantitative comparisons to representative flownet-based and transformer-only baselines. The results demonstrate competitive or improved performance in standard translation and rotation metrics on the majority of sequences. We also discuss failure modes (e.g., low-texture regions and highly dynamic scenes) and limitations of using monocular inputs alone. Future work will explore multi-sensor integration (e.g., GNSS, IMU or stereo) to further enhance long-term consistency and produce more reliable navigation solutions.
Keywords: FlowVoT; Monocular Visual Odometry; Optical Flow; FlowNet; Vision Transformer; Pose Estimation; Autonomous Navigation.

Low Earth Orbit (LEO) satellites offer promising prospects for advancing navigation and positioning technologies. Nevertheless, the Two-Line Element (TLE) ephemeris, which serves as a pivotal data source for LEO-based positioning, is constrained by its infrequent update intervals, thereby impeding real-time positioning performance. To mitigate this challenge, this study introduces a physics-informed neural network (PINN) model tailored for TLE ephemeris prediction. The investigation initiates with an analysis of the TLE ephemeris data, revealing that angular momentum can be effectively approximated as conserved over a one-year timescale. Building upon this empirical observation, a physics-informed Long Short-Term Memory (PI-LSTM) model architecture is developed, integrating angular momentum conservation principles as physical constraints to improve TLE parameter prediction. Experiments demonstrate that the PI-LSTM model achieves a 19% improvement in TLE parameter prediction accuracy relative to conventional LSTM and Transformer models. Moreover, comparative experiments are conducted employing four distinct strategies: TLE interpolation, LSTM-aided, Transformer-aided, and the proposed PI-LSTM model. The results indicate that the proposed approach not only significantly surpasses the other three methods in prediction accuracy but also maintains computational efficiency compared with the deep learning-based alternatives.

Precise orbit and baseline determination is a fundamental prerequisite for formation-flying low Earth orbit (LEO) satellites engaged in interferometric synthetic aperture radar (InSAR) missions. The SuperView Neo2-03/04 (SVN2-03/04) satellite pair operates in a close formation and is equipped with spaceborne GNSS receivers capable of tracking dual-frequency GPS and BeiDou (BDS) signals. This configuration provides a valuable opportunity to assess the performance of multi-constellation GNSS-based precise orbit determination (POD) under real mission conditions.
This study evaluates the precise orbit determination (POD) performance of the SVN2-03/04 satellite pair using dual-frequency observations from GPS-only, BDS-only, and combined GPS+BDS systems. The data were collected over an 18-day period from March 21 to April 7, 2025 (corresponding to Day of Year 80–96). Both kinematic and reduced-dynamic orbit determination strategies are applied. Orbit quality is assessed through carrier phase residual analysis, overlapping arc consistency, and comparisons between reduced-dynamic and kinematic solutions.
The results indicate that the spaceborne GNSS observations maintain stable measurement quality, with carrier phase residuals remaining at the millimeter level across all processing strategies. Relative to single-constellation solutions, the combined GPS+BDS approach yields consistently improved internal orbit consistency. Overlapping arc analyses show average three-dimensional RMS values of approximately 7.62 mm for BDS-only solutions, 14.50 mm for GPS-only solutions, and 2.21 mm for GPS+BDS combined solutions. Further comparisons between reduced-dynamic and kinematic orbits suggest enhanced robustness when multi-constellation observations are employed.
Overall, the in-orbit results demonstrate that multi-constellation GNSS observations can effectively improve the internal consistency and stability of precise orbit solutions for formation-flying LEO satellites, satisfying the centimeter-level orbit accuracy requirements of the SVN2-03/04 interferometric mission. The findings offer practical insight for the design and operation of future LEO formation missions based on multi-GNSS POD.

In disaster-response and urban operations, Unmanned Aerial Vehicles (UAVs) often experience degraded or intermittent Global Navigation Satellite System (GNSS) positioning due to multipath, blockage, or interference, which can destabilize navigation and coordination. While many cooperative localization approaches focus on estimation and fusion for a given team configuration, and some geometry-aware methods use Dilution of Precision (DOP) for reference selection or centralized planning, relatively limited work integrates PDOP-guided distributed reconfiguration for cooperative self-localization under mission and separation constraints. We propose cooperative localization with PDOP-guided distributed optimization, where a UAV team combines inertial sensing with inter-UAV ranging while maintaining mission trajectory tracking and enforcing collision avoidance through minimum-separation constraints. A multi-UAV simulation environment is developed with mission trajectories, ground-truth motion, and sensor models including an Inertial Measurement Unit (IMU) and time-varying GNSS quality (e.g., increased noise and temporary outages). When GNSS becomes unreliable, UAVs exchange peer-to-peer range measurements to enable cooperative multilateration, producing position estimates with an associated geometric-quality characterization. PDOP is used as an interpretable driver within a constrained cooperative cost function, and the team formation is adjusted via distributed iterative updates in which selected UAVs act as mobile anchors to improve the ranging geometry for one or more target UAVs. The resulting GNSS, IMU, and cooperative multilateration information is fused using a Kalman Filter (KF) while incorporating multilateration uncertainty. Preliminary simulations indicate improved robustness under degraded GNSS, with up to ~30% reduction in mean PDOP and up to ~15% reduction in mean 3D RMSE relative to non-cooperative GNSS/IMU estimation; performance depends on scenario constraints and anchor selection/update ordering. Worst-case 3D errors are on the order of several metres (~5–10 m) in the tested scenarios.

Recent breakthroughs in the field of Large Language Models (LLMs) have enabled the emergence of highly capable Artificial Intelligence (AI) agents. Initially designed to process natural language, the latest research contributions have introduced various techniques (function calling, prompt engineering, web-research, retrieval-augmented generation…) to unlock LLM’s potential toward versatile assistants, able to perform a wide range of tasks. Today, the efficiency of AI agents makes them valuable tools for companies worldwide, particularly for software development teams by helping them to accelerate their development cycles and reducing their time to market, which represents a considerable asset in competitive industries.

Capitalizing on the remarkable coding skills of LLMs, especially in Python language as demonstrated by several benchmarks, the Skydel AI project proposes a specialized coding assistant to enhance the user experience of the Skydel simulator. Accessible through a web application with an intuitive user interface, the project values the advanced capabilities of LLMs in debugging, suggesting, and generating code through an innovative retrieval-augmented generation (RAG) pipeline to support the simulation needs of Skydel users. Our approach aims to help users fully exploit the extensive functionalities offered by the Skydel Python API, by introducing an assistant with dual capabilities: (1) suggest relevant commands and custom configurations, and (2) generate ready-to-run Skydel simulation scripts. This paper details the methodology applied and presents the results of the RAG evaluation , highlighting the quality of the assistant.

Irregular variations in ionospheric electron density severely affect trans-ionospheric radio frequency signals and constitute one of the major error sources for Global Navigation Satellite System (GNSS). The refraction experienced by GNSS signals as they propagate through the ionospheric plasma is referred to as ionospheric delay. Currently, GNSS receiver predominantly employs dual-frequency observations to calculate the Total Electron Content (TEC), which substantially mitigates the effects of ionospheric delay. However, due to the non-ideal characteristics of hardware channels within the satellites and receivers, the Differential Code Biases (DCBs) is introduced. The analysis center of International GNSS Service (IGS) provides global ionospheric TEC maps as well as estimated DCB values for GNSS satellites and IGS ground stations to support ionospheric monitoring. However, due to the uneven global distribution and insufficient real-time capability of ground stations, ionospheric monitoring in certain areas still faces challenges.
In recent years, ionospheric monitoring has effectively addressed the limitations of traditional ground stations utilizing mobile platforms, e.g., mobile phones, vessels, and subways. With the development of the low-altitude economy, the widespread use of Unmanned Aerial Vehicles (UAVs) provides a new option for ionospheric monitoring. The potential factors influencing airborne ionospheric monitoring is investigated by combining UAV with ground stations. Ionospheric monitoring data collected from UAV and ground station during April 5th to 7th 2023 is used to explore the influencing factors.
Precise Point Positioning (PPP) technology is employed to determine the real-time flight trajectory of the UAV and the position of the ground station, along with the elevation and azimuth angles of visible satellites. The true TEC values at the Ionospheric Pierce Points (IPPs) for both the UAV and the ground station are obtained from Center for Orbit Determination in Europe (CODE) products. Pseudorange and carrier phase observations are acquired from the Receiver Independent Exchange Format (RINEX) files provided by the UAV and the ground station. The observations are then used to generate the geometry-free combination measurements for comparative analysis.
The experimental results indicate that platform motion introduces errors into the measurements. Specifically, the standard deviation of the P4 measurements from the UAV increases compared to that from the ground station. This increase subsequently leads to larger errors in the calculation of the DCBs from UAV and the predicted TEC values, ultimately affecting the accuracy of ionospheric monitoring based on the airborne platform.

As maritime operations become increasingly dynamic and safety-critical, there is a growing need for training systems that respond more effectively to the individual needs of seafarers. This study examines how artificial intelligence (AI) can enhance feedback within maritime education by comparing AI-generated feedback with traditional rule-based explanations following a COLREGs knowledge assessment. Forty-nine active seafarers reviewed the feedback they received and evaluated it across several instructional dimensions, including clarity, sufficiency, correctness, usefulness, adaptiveness, and motivational impact.

All AI feedback was prepared in advance using structured prompts that incorporated the correct answers and common error patterns, ensuring consistency and instructional accuracy. Participants completed the COLREGs assessment and then rated both the traditional expert-written explanations and the AI-generated feedback. Statistical analysis demonstrated clear differences between the two approaches. Seafarers consistently rated the AI-generated feedback higher, describing it as clearer, more relevant, and more helpful for understanding mistakes. They noted that it highlighted specific weaknesses more effectively and provided explanations that were easier to interpret and apply.

Open-ended comments also indicated strong acceptance of AI-supported feedback. Participants expressed that AI explanations felt more personalised, more engaging, and more aligned with their actual learning needs than the generic rule-based responses typically used in current training systems. Some noted that the AI feedback presented information in a way that supported reflection and reinforced the correct application of COLREGs.

The findings suggest that AI-enhanced feedback has significant potential to strengthen maritime training by delivering tailored guidance, improving the clarity of feedback, and reducing dependence on static, one-size-fits-all methods. While this study focuses on learner perceptions rather than long-term learning outcomes, it provides strong evidence that seafarers view AI-generated feedback positively and see it as a valuable component of future maritime training systems.

Accurate localization of individuals requesting emergency assistance is essential for rapid response and reduced rescue time. Although the Global Navigation Satellite System (GNSS) provides reliable navigation performance in open-sky environments, its accuracy deteriorates significantly in urban canyons, tunnels, and underground spaces due to signal blockage and multipath signals. To ensure location availability and accuracy in such GNSS shadow areas, alternative infrastructure or sensor-based positioning methods are required. This study proposes a map-matching system that uses Pedestrian Dead Reckoning (PDR) constraints computed from smartwatch inertial data together with a numerical map to estimate pedestrian trajectories. The proposed system consists of three key components. The first component is the smartwatch device, which samples raw inertial data from its Inertial Measurement Unit (IMU) at fixed intervals and transmits it to a positioning/monitoring server in real time via the commercial network. This component provides essential input data for computing PDR information based on wrist-mounted motion. The second component, corresponding to PDR, and the third component, corresponding to map-matching, operate on the positioning/monitoring server. The PDR processing module performs preprocessing of the inertial signals and applies a Human Activity Recognition (HAR) approach to distinguish between gait and non-gait segments. Step detection, stride length estimation, and turning detection are then performed to generate PDR information. This information is stored as a time-ordered sequence of distance and turning states. The subsequent map-matching module models the pedestrian road network as a node–edge graph and iteratively removes candidate paths that are inconsistent with the temporal PDR constraints. Through this process, the trajectory that satisfies all PDR constraints is selected as the final estimated path. The proposed system enables continuous pedestrian path tracking even in GNSS shadow areas. It does not require additional infrastructure and allows immediate access to estimated path results through direct data communication between the smartwatch and the processing server. Experiments conducted in actual urban environments considered various walking conditions, including gait and non-gait segments, variations in gait patterns, and turning. The results demonstrated that the estimated paths closely aligned with the true trajectories. These findings show the potential of the proposed system for applications in public safety, emergency rescue, and other location-based services.

A new approach in geodetic reference frames transformation is proposed in this article for swift establishment of geodetic networks accurately. In one hand, GNSS surveys provide us a set of points coordinates expressed in global systems with millimeter accuracy, in other hand, the same set of points formally determined by classical technics have a local coordinate quietly less accurate (around few centimeters).
In order to fix these discrepancies in localization, some transformation formalisms have been introduced since many decades to allow such local geodetic networks of every country to express their coordinates locally with accepted precision.
Despite the efforts consented in developing and validating Helmert 3D-similitude with its two variants (Molodensky-Badedas, Bursa Wolf), similitude 9 parameters, Multiple Regressions Equations and many other models relatively early developed such as Vanicek Featherstone which give an accuracy around 9 cm on Δx, 1.08 m on Δy and 1.45 m on Δz
, the problem remain non completely or partially resolved specially for a country with huge non uniform and distorted areas like Algeria where the precision and signification of transformation parameters are nowadays weak. Taking in account these troubles, I have introduced and tested the relatively new approach called quaternion method based on dual quaternion imaginary and real numbers as follows Q = (x1i + x2j + x3k) + w0.

To validate the method, a set of sufficient double points coordinates served to initiate the dual numbers and then to proceed on their iterations in the design matrix. To resolve the unknown so called transformation parameters equation and determine them, the number of double points have been changed for carrying out a comparative results . Using the known coordinates of the double points, we perform thus the Implementation of the dual quaternion algorithm for 3D similarity-based coordinates transformation between Algerian local geodetic datum and GNSS frame WGS84. The results confirmed that the coordinates transformed by the dual quaternion algorithm are in average agreement with the measured coordinates, with precision and accuracy levels of about 0.26 m in planimetric and 0.78 m in up.

LunaNet is a proposed lunar communication and navigation network designed to provide robust, interoperable, and scalable services for future exploration missions. Developed through international collaboration among NASA, ESA, and JAXA, LunaNet aims to establish an architecture analogous to terrestrial Internet and GNSS systems, enabling high-data-rate communications, precise navigation, and the relay of science data across the lunar environment. Within this framework, the Lunar Augmented Navigation Service (LANS) provides scalable and interoperable positioning, navigation, and timing (PNT) capabilities. The Augmented Forward Signal (AFS), a standardized navigation signal structure defined for LANS, supports GNSS-like services through signals broadcast by multiple LunaNet Service Provider (LNSP) satellites in lunar orbit.

To support LNSP satellite development and enable end-to-end evaluation of AFS waveform generation and receiver processing algorithms, we previously developed an AFS simulator and a receiver using software-defined radio (SDR). However, at that time, an S-band RF frontend designated for AFS reception was not yet available, and the system was evaluated using L-band signals compatible with conventional GNSS hardware. In this study, we assess the NTLab NT1066 as a candidate S-band frontend for LunaNet AFS. The NT1066 is a four-channel multi-band GNSS frontend IC with one channel designed explicitly for S-band operation. Because it provides 2-bit ADC samples, it can be seamlessly integrated into the existing receiver architecture, initially developed for 2-bit GNSS L-band frontends.

In addition, we investigate the applicability of general-purpose SDR hardware, including USRP, LimeSDR, and bladeRF, not only for signal transmission from the simulator but also as configurable receiver frontends. These devices typically provide baseband signals with 12-bit ADC quantization, whereas our SDR-based receiver requires 2-bit samples to enable efficient bit-wise correlation processing using SIMD instructions. To bridge this mismatch, we implement a digital automatic gain control (AGC) stage that adaptively compresses the 12-bit baseband samples into 2-bit quantized values while preserving correlation performance.

By integrating these components, we establish a fully functional SDR-based testbed capable of evaluating both the LANS AFS signal structure and its compatibility with the S-band frequency allocation specified for LunaNet. This testbed provides a practical and flexible platform for validating the entire AFS signal chain, from waveform generation to real-time receiver performance, in preparation for future lunar navigation services.

This research was supported by JAXA SSF Program Japan Grant Number JPJXSSF24ME13001.

This paper shows how global navigation satellite services (GNSS) can play a valuable role beyond their intended use, namely in efficient spectrum management. When different radio applications share the same spectrum, the separation by attenuating material is a way to mitigate potential interference. The indoor restriction for WLAN devices in 5150-5350 MHz is an example for a regulatory measure that aims at having WLAN devices operating in an environment that provides sufficient attenuation to enable sharing with other services. However, the concept of “indoor” is ambiguous and encompasses diverse radio environments. This paper proposes a novel method of how an attenuating environment can be automatically detected without user interaction. Instead of detecting an indoor location, we are directly looking for a detection of an attenuating environment. The basic idea is that GNSS signals can be received practically everywhere on earth where there is a view to the sky. Where these signals are attenuated, the receiving device is assumed to be in an attenuating environment, which provides some level of shielding to the environment. In order to characterize such environment, we evaluate the detectable GNSS satellites and their carrier-to-noise density. Example measurements show that GNSS raw data can help to distinguish between low-attenuating locations and higher-attenuating locations. These measurements were conducted with GNSS receiver in an off-the-shelf Android tablet in order to show the feasibility of the approach, without a need for specialized equipment. The results highlight that GNSS data, originally designed for positioning and navigation, can be repurposed in innovative ways that go beyond its traditional use.

Global Navigation Satellite Systems (GNSS) have become the indispensable backbone of modern critical infrastructure. However, the inherent technical vulnerability of GNSS signals (e.g., extremely weak signal power, often -20dB below the thermal noise floor for civil signals) has created a profound asymmetry between the low-cost, readily available malicious threats (Jamming/Spoofing) and the complexity of defense. This asymmetry is further exacerbated by escalating geopolitical tensions, which have transformed GNSS disruption into a persistent, systemic threat, undermining Positioning, Navigation, and Timing (PNT) services.

This paper addresses the critical question of how national and international legal/institutional frameworks can effectively resolve this technical asymmetry and secure national PNT resilience.

First, we analyze the core technical capabilities: the offensive ‘spear’ (Jamming/Spoofing techniques including Continuous Wave (CW), Noise, Sweep for Carrier-to-Noise Density Ratio (C/N0) degradation and low-cost signal simulators) and the defensive ‘shield’ (Anti-Jamming techniques like Controlled Radiation Pattern Antenna (CRPA) Beamforming and Anti-Spoofing methods such as Galileo’s Open Service Navigation Message Authentication (OSNMA) and Receiver Autonomous Integrity Monitoring (RAIM)).

Second, we analyze how major legal frameworks (International Telecommunication Union (ITU) Harmful Interference control, International Civil Aviation Organization (ICAO) / International Maritime Organization (IMO) safety regulations) establish the foundation for technical protection, particularly by mandating minimum performance requirements for safety-critical GNSS equipment.

Finally, we conduct a comparative analysis of PNT resilience strategies across three distinct models:

1. European Union (Galileo): Security by Design, integrating legal security provisions with high-assurance technical solutions like OSNMA and Public Regulated Service (PRS) from the system’s inception.

2. United States (GPS): Market Regulation and Alternative PNT Legal Support, utilizing Federal Communications Commission (FCC) regulations to suppress threat sources and providing statutory support for enhanced Loran (eLoran).

3. Republic of Korea (Security Threat): Nation-State Radio Frequency Interference (RFI) Countermeasures, focusing on rapid national response systems and the legal basis for Korea Augmentation Satellite System (KASS) and eLoran deployment against continuous, external threats.

The analysis reveals that a strong, integrated legal framework is crucial to translating policy into a measurable, performance-based technical ‘shield.’ This study concludes by proposing policy recommendations for unified PNT resilience governance, specifically advocating for the international or domestic legal obligation to equip safety-critical PNT devices with essential Anti-Spoofing technology (OSNMA-equivalent), and promoting a global harmonization of performance standards to ensure enduring PNT resilience.

Robust GNSS tracking on embedded software-defined radio (SDR) platforms must sustain carrier/code lock under high dynamics and weak-signal conditions while meeting strict real-time constraints. Kalman-filter-based tracking is attractive since it couples tracking-loop estimation with an explicit dynamic state model; however, the per-epoch prediction and measurement update can be computationally prohibitive on low-power processors, particularly for multi-channel operation. Lookup-table (LUT) Kalman filter (KF) tracking reduces steady-state cost through precomputed gains; however, fixed transient tuning often assumes benign initial conditions, which can result in slow convergence or loss of lock when large initial Doppler and carrier-phase errors occur during acquisition handover.

This paper proposes a hybrid tracking architecture implemented on an ADI Jupiter SDR running FreeRTOS on an ARM Cortex-A53. During pull-in and loss-of-lock recovery, the receiver employs a discrete-time Kalman filter whose arithmetic cost is reduced by explicitly exploiting model structure: structurally zero terms in the state-transition and measurement updates are removed (zero-skipping), constant patterns are algebraically expanded into a minimal set of multiply–accumulate operations, and redundant computations are avoided by leveraging the symmetry of the covariance matrix. After convergence, the receiver transitions to an adaptive full LUT-KF mode, where the LUT index is updated online via a loop-bandwidth control algorithm (LBCA) driven by discriminator statistics, enabling the loop to widen for robustness and narrow for precision as conditions vary. A bandwidth-matched handover aligns equivalent loop dynamics before switching to suppress transients and preserve carrier-phase continuity.

The proposed approach is evaluated in two complementary scenarios: (i) high-dynamic simulations that stress pull-in and tracking robustness under rapid Doppler and jerk variations, and (ii) real-signal dynamic vehicle tests using GPS L1 C/A over-the-air measurements. Results show improved pull-in robustness and reduced time-to-lock compared with fixed-transient LUT approaches, while maintaining stable steady-state tracking. Profiling on the target platform confirms that the structure-exploiting discrete KF provides additional real-time headroom during pull-in, and the LUT mode achieves a low steady-state processing load, making it suitable for embedded multi-channel tracking while leaving a compute budget for navigation, logging, and integrity-monitoring tasks.

The advancement of global Precise Point Positioning (PPP) services is pivotal for enhancing the performance and robustness of next-generation satellite navigation. However, unlike standardized Satellite-Based Augmentation Systems (SBAS), the current proliferation of PPP solutions features disparate service architectures, signal structures, and data formats, which compatibility and interoperability. This research directly addresses this landscape by investigating viable pathways toward PPP interoperability. Our study conducts a comprehensive analysis spanning from physical signal characteristics through message content to user-level algorithms and receiver channel processing. Based on this, we design and propose concrete technical pathways for interoperability at each layer, evaluating them against performance, complexity, and backward compatibility. Finally, the paper provides actionable recommendations for service providers and standardization bodies to foster compatible and interoperable global PPP services, ultimately improving accessibility and delivering a seamless user experience worldwide.
Based on a layered technical framework, our study conducts a comprehensive analysis from the system level down to the user terminal. (1) At the system level, we analyze the service coverage areas and dissemination strategies of current global navigation satellite systems. (2) At the signal level, the compatibility and interoperability characteristics of major PPP signals are investigated, focusing on physical parameters such as center frequency, bandwidth, and equivalent carrier-to-noise density ratio. (3) At the message level, the combined usage and interoperability of existing precise correction formats and dissemination protocols are examined. (4) Finally, at the user terminal level, an integrated architecture for jointly receiving, processing, and utilizing multi-system signals and messages is analyzed and proposed.

Abstract:
In many developing regions, local governance and sustainable development initiatives face complex challenges related to spatial planning, rural accessibility, population mobility, resource allocation, and the availability of accurate, timely data for decision-making. Navigation technologies, particularly Global Navigation Satellite Systems (GNSS) and Positioning, Navigation and Timing (PNT) solutions, have emerged as essential tools to address these challenges, providing precise location, timing, and coordination in diverse and often remote environments.

This paper explores how GNSS, when combined with artificial intelligence (AI) and geospatial data analytics, can significantly enhance local governance and improve the implementation of community-based development projects. GNSS provides precise information on the location of communities, infrastructures, and intervention areas, answering the critical question of where. AI acts as an intelligent data search and navigation engine, analyzing, organizing, and interpreting vast volumes of GNSS and field-collected data. This process enables decision-makers to identify priorities, allocate resources efficiently, and respond to emerging needs, much like a web search engine, but tailored to development contexts.

Drawing on extensive field experiences conducted by civil society organizations, this contribution highlights practical applications of GNSS in participatory mapping, territorial planning, monitoring project logistics, and assessing the impacts of development initiatives. The integration of AI facilitates optimization of team coordination, evidence-based prioritization of interventions, and improved transparency and accountability in project execution. These technologies allow local authorities and community stakeholders to make timely, accurate, and inclusive decisions that enhance resilience and social cohesion.

Special emphasis is placed on resource-constrained environments, where limited digital infrastructure and budgetary constraints require innovative yet pragmatic approaches. These include the use of low-cost mobile devices, multi-sensor navigation solutions, AI-powered analytics adapted to local conditions, and community training programs. Such interventions demonstrate how technology can overcome traditional barriers to development, enabling more equitable access to decision-making processes and empowering local communities.

Beyond the technological dimension, human and institutional factors are essential for sustainable adoption. Leadership, capacity building, and collaboration among community members, local authorities, and technical experts are critical to ensure that GNSS and AI tools are effectively utilized and maintained. Participatory engagement strengthens ownership and ensures that technological interventions respond to actual local needs.

This paper presents a framework illustrating the synergistic potential of GNSS and AI to support sustainable development goals, enhance local governance, and foster social innovation. Lessons learned, best practices, and recommendations for scaling these technologies in similar contexts are discussed. By combining precise navigation data with intelligent analytics, stakeholders can plan, monitor, and evaluate projects more effectively, aligning with the theme of ENC 2026, “Navigating the Future: Seamless, Smarter, More Performant.”

Traditional ship weather routing systems usually output a single deterministic route recommendation that inadequately represents the inherent uncertainties in weather forecasts and vessel performance predictions. This prevents meaningful decision support. Single-solution approaches eliminate the ability to evaluate differences between competing operational priorities such as fuel efficiency, schedule reliability, and weather avoidance. This research addresses these limitations by developing a comprehensive stochastic optimization framework that transforms weather routing from prescriptive automation into genuine on-board decision support.
The proposed framework integrates ensemble weather forecasts with neural network-based ship performance models and hierarchical optimization algorithms to generate multiple strategic route alternatives with probabilistic constraints. A hybrid approach coupling A* global path planning with Stochastic Model Predictive Control (SMPC) enables simultaneous consideration of long-term strategic planning and short-term tactical execution. The system quantifies weather forecast uncertainties through detailed analysis, capturing temporal uncertainty evolution from 24 to 168 hours ahead. Validation across 21 transatlantic voyage scenarios demonstrates that the multi-alternative framework provides ship operators with diverse routing options according to different operational priorities. The system generates candidate routes optimized for fuel consumption predictabiluty (361-426 tonnes), maximum schedule reliability (±6-8 hour ETA variance), and robust performance across weather scenarios (85-93 % scenario feasibility). Real-time computational efficiency enables on-board implementation with re-planning intervals of 6-8 hours. This approach redefines Decision Support Systems (DSS) by establishing that meaningful decision support cannot exist without alternatives. By providing probabilistic voyage outcome distributions rather than deterministic point estimates, the system enables informed risk-aware decisions balancing competing operational objectives. Implementation recommendations differentiate between coastal voyages (24-72 hour horizons), mid-range transoceanic passages (5-10 days), and extended voyages requiring more sophisticated uncertainty management.

Autonomous estimation of rotational motion for non-cooperative space objects is essential for on-orbit servicing, inspection, and debris monitoring missions operating in GNSS-denied environments. While deep learning methods have demonstrated impressive performance, they demand substantial computational resources and lack interpretability—critical limitations for resource-constrained spacecraft where algorithm transparency and fault tolerance are paramount.
This paper investigates traditional computer vision methods as computationally efficient and interpretable alternatives for estimating the principal rotation axis and angular velocity of tumbling objects from monocular image sequences. We systematically compare two classical approaches: (1) feature-based tracking using ORB and SIFT detectors, where rotation parameters are extracted through Essential Matrix decomposition and geometric analysis of tracked point correspondences; and (2) dense optical flow using the Lucas-Kanade method, which infers rotational motion directly from pixel-level displacement fields.
Both methods are evaluated on the SHIRT (Satellite Hardware-In-the-loop Rendezvous Trajectories) dataset, providing sequential images with precise ground truth for single-axis and complex two-axis tumbling scenarios. Performance analysis reveals that feature-based methods excel in rotation axis estimation when sufficient texture is available, while optical flow provides more robust angular velocity estimates under challenging illumination variations. Single-axis rotation presents unique observability challenges due to limited geometric variation, whereas two-axis tumbling offers richer motion information at increased computational cost.
This work establishes performance bounds for embedded-friendly vision algorithms, providing interpretable baselines for future hybrid architectures. The findings directly support the development of lightweight, sensor-minimal relative navigation systems for autonomous space missions in GNSS-denied environments.

In addition to real-time field test, two widely used methods for testing GNSS devices and systems exist today. Recording and playback of real signals in actual environments, and simulation/emulation of these in a lab environment. Each methodology plays its part, but neither encompasses it all.
GNSS record and playback systems (RPS), enable a real RF environment to be captured, stored and repeatedly replayed to a device under test (DUT). This allows field-observed receiver behaviours to be replayed, repeatedly in the lab.
But until now, there was no way to understand the composition of the RF recording, so there was no way to know which elements in the GNSS signal environment produced what response in the DUT. That level of insight could only be gained by using controllable, synthetic signals generated by a GNSS simulator.
While simulators are today capable of great deal of realism, it can be time-consuming to set up highly realistic scenarios, and no simulator can fully replicate the full richness of a specific real-world environment. During this presentation I wish to provide insight into the development and performance of a novel, patent-pending solution that analyses an RF recording and decomposes it into individual GNSS signals, including line-of-sight (LOS) signals, multipath reflections, and the key parameters of each, and uses the data to create a fully-configured scenario for use in a PNT simulator.
It effectively brings the actual, real signals into the lab, but with the full knowledge, flexibility, control and other benefits of a GNSS simulator, such as adding in other forms of signal, or interference which may be very rare in the real world, but which need to be trialled, for example; jamming and spoofing.
It provides the receiver developer with truth data for the real-world signals, allowing them to compare their device’s measurements against what was present, to enhance their measurement performance.
The benefits of this are that realistic signals (the ones that the device being developed will, in the end, be exposed to in operation) can be introduced to the development test cycle at any stage, in a controlled way. It allows more comprehensive field test coverage from the same field test investment. This ensures designs can be made robust from early development stages and potential issues found much sooner in the process, saving time and money versus discovering them at integration or even deployment stage.

Global Navigation Satellite Systems (GNSS) remain the primary source of absolute positioning and velocity for many safety critical and autonomous applications. However, the growing prevalence of intentional and unintentional GNSS (Global Navigation Satellite System) interference—particularly jamming and spoofing—has highlighted the need for resilient alternative and aiding navigation sources. The Honeywell Radar Velocity System (HRVS) is presented as a robust velocity aiding sensor, designed to maintain navigation performance in contested or degraded environments.
HRVS provides direct, drift free velocity measurements using Doppler principle, independent of external infrastructure, environmental or lighting conditions. Unlike inertial sensors, whose velocity estimates degrade over time due to bias accumulation, HRVS delivers absolute velocity observability relative to the surrounding environment. When tightly coupled with an inertial navigation system (INS), HRVS stabilizes velocity, roll and pitch drift, slowing position drift during GNSS outages and enabling rapid recovery once GNSS becomes available again.
While HRVS aided inertial navigation represents a highly effective solution, it remains a fundamentally relative navigation approach. Residual position drift arises from velocity integration combined with heading errors driven primarily by gyro bias. To address this limitation, a terrain based map matching extension is introduced, enabling absolute navigation by correlating measured terrain profile with terrain elevation data. Terrain referenced navigation provides absolute position measurements with typical accuracies on the order of ±50–100 m; however, its performance is constrained in regions with low terrain variability.
To meet more demanding absolute accuracy requirements, down to approximately 10 m, the navigation architecture is further extended with vision based map matching aiding. Vision based techniques provide high accuracy absolute position fixes by correlating onboard imagery with pre mapped features but benefit strongly from stable motion prediction and reduced search space. HRVS augments both terrain aided and vision based navigation by delivering reliable short to mid term navigation solution stability. This synergy enables the use of significantly lower cost inertial measurement units while maintaining high navigation performance.
Additionally, since HRVS measurements can be leveraged to measure height above ground level (AGL), the system is particularly suitable for platforms with low-level flight trajectory. The presented multi sensor navigation architecture demonstrates a scalable and resilient approach to assured navigation in environments where continuous GNSS availability cannot be assumed.

Reliable positioning for land vehicular navigation in urban environments is a challenging issue for GNSS due to severe signal multipath interference. To address this challenge, integrating GNSS with inertial measurement units (IMUs) has been proven to be an effective solution, especially when augmented with motion constraints, observation weighting, or other advanced algorithms. However, at the sensor fusion stage, conventional filtering designs suffer from inherent limitations, primarily because real-world measurement noise is rarely perfectly Gaussian. Addressing this gap, this study focuses on enhancing positioning performance through a factor graph optimization (FGO) sensor fusion framework. Crucially, signal quality monitoring (SQM) is incorporated to detect signal anomalies, providing reliability metrics that serve as the basis for measurements weighting within the FGO, effectively compensating for the absence of multipath integrity processing in standard receivers.
Common integration strategies, typically based on filtering techniques such as the Kalman Filter (KF) and its variants, have served as the cornerstone of sensor fusion. However, these methods suffer from inherent limitations, including susceptibility to non-Gaussian noise with high uncertainty and recursive estimation based solely on the immediate past. In contrast to recursive filtering, FGO adopts a global optimization perspective, formulating the GNSS/INS integration as a joint nonlinear least-squares problem. This framework allows for the iterative refinement of past states through re-linearization, effectively mitigating the accumulation of linearization errors inherent in KF. It is important to note that in multipath environments, degraded GNSS measurements can easily impair optimization performance, particularly when outliers are indistinguishable from valid observations. To overcome this restriction, this study also incorporates signal quality monitoring (SQM) to detect signal distortions at the tracking level. SQM utilizes an additional pair of early and late correlators to evaluate signal reliability based on metrics such as correlator peak asymmetry and carrier to noise ratio (C/N0) fluctuations. These metrics are integrated into the covariance matrix of the GNSS factors, down-weighting measurements affected by multipath to rectify the inaccurate weighting inherent in conventional models. Consequently, this research adopts FGO enhanced with SQM as the core of the sensor fusion algorithm.
Moreover, bridges, buildings, or other obstacles frequently induce multipath interference, non-line-of-sight (NLOS), and signal obstruction in urban environments. These conditions often disrupt GNSS tracking loops, resulting in signal loss of lock and data outages. Therefore, this study implements a deep integration scheme to ensure stable GNSS signal tracking, where the estimated velocity information is fed back into the numerically controlled oscillator (NCO) of the GNSS receiver’s tracking loops.
To validate the effectiveness of the proposed framework, real-world dynamic vehicular field experiments were conducted in signal-degraded scenarios. The performance of the proposed algorithm was compared with that of the standard KF and the FGO scheme without SQM. Finally, results show that while FGO offers better global consistency than recursive filtering, the addition of SQM provides critical information for measurement weighting in the presence of severe multipath interference. Finally, the proposed FGO-SQM framework achieves the highest positioning accuracy among the tested methods, proving its effectiveness for navigation in the experimental scenarios.

Forests play an essential ecological and economic role across the word. Forest data serve multiple purposes, including estimating national forest volumes, assessing regional biomass, evaluating carbon sequestration, and analyzing water balance across different ecosystems. Additionally, with the rapid advancement of digital technologies, forest data are increasingly used within the framework of precision forestry, enabling management decisions to be made at the individual tree level.

To enhance knowledge about forests and support sustainable forest management and right decision marking, a wide range of tools and technologies have been developed and implemented over the past decade. These include advanced field instruments such as digital tree caliper and hypsometer, as well as advanced remote sensing technologies including differential Global Navigation Satellite Systems (GNSS), Simultaneous Localization and Mapping Technologies based on laser scanning and camera sensors, and pseudo-satellite technologies such as ultra-wideband (UWB). Terrestrial Laser Scanning (TLS) and Mobile Laser Scanning (MLS) have been extensively studied and applied to obtain high-precision reference data. However, the overall progress over the last 10-15 years has been relatively modest. The primary challenge lies in the strong dependences of these methods on the quality of the employed laser scanning sensors. Higher accuracy can be achieved with dense and high-precision scanners, but this inevitably results in increased system cost, power consumption, data collection complexity, and intensive post-processing efforts. As a result, there is an emerging and growing demand in both research and industry for an optimized solution that balances all these aspects simultaneously.

UWB technology has recently emerged as one of the most promising approaches for achieving this balance. Studies have demonstrated that UWB systems offer an exceptional combination of advantages: compact device size (< 3×3×1 cm3), low weight (< 100 g), minimal power consumption (< 0.5 w), affordable system cost (< 1000 euros), fast deployment (15-20 minutes), adequate measurement accuracy (10-30 cm), real-time edge-computing capability, user-friendly operation (no professional training required), drift-free localization, and scalability (typically coverage distance exceeding 40 m). All these achievements make UWB technology highly suitable for both independent and integrated applications in precision forestry, such as tracking drones under dense canopies or measuring tree position, height, diameter at breast height (DBH) and species when combined with traditional field instruments such as a tree caliper.

This study focuses on evaluating the performance metrics of a UWB data-driven method for precision and efficiency field surveying, emphasizing its commercial potential in precision forestry. The results highlight that UWB-based methods can significantly enhance operational efficiency, reduce system cost and deployment time, and provide a practical path toward scalable and intelligent forest monitoring systems. Consequently, UWB technology represents a promising step forward in bridging the gap between academic research and the large-scale commercialization of precision forestry solutions.

High-precision real-time satellite clock products are essential for Low Earth Orbit (LEO) satellite–augmented Positioning, Navigation, and Timing (PNT) services. The inevitable delays in observation data transmission and clock estimation make LEO satellite clock prediction indispensable to mitigate latency and deliver real-time satellite clocks to users. However, LEO satellite clocks exhibit complex periodic behavior posing substantial challenges for prediction, particularly over medium- and long-term. Compared with traditional polynomial–periodic models, deep learning techniques are better suited for extracting useful information from historical clock series. Therefore, this study introduces a wavelet-enhanced Long Short-Term Memory (LSTM) model for predicting LEO satellite clocks, aiming to provide users with high-precision real-time satellite clock products in case of relatively long delays, say minutes to tens of minutes.
The wavelet-LSTM model proposed in this study is implemented as follows. First, the original clock series is decomposed into six levels using the Maximal Overlap Discrete Wavelet Transform (MODWT) with the Daubechies-4 (db4) mother wavelet, yielding six detail components and one approximation component that effectively isolate different frequency structures. Second, each decomposed component is independently modeled using an LSTM network, allowing the model to capture the distinct temporal patterns within each frequency band. Finally, the predicted sub-series are reconstructed to obtain the overall clock prediction, and the prediction accuracy is evaluated by comparing the predicted clocks with the post-processed precise clocks using the Root Mean Squared Error (RMSE).
The proposed method was evaluated using Sentinel-3B onboard GNSS observations from Day Of Year (DOY) 226–252 in 2018. The satellite clocks were first estimated and then predicted, with prediction terms of 5, 10, 30, and 60 min. The results indicate that the wavelet-LSTM approach effectively captures the complex temporal characteristics of LEO satellite clocks and achieves superior prediction accuracy across all prediction terms. Compared with the traditional polynomial–periodic model, the proposed method improves prediction accuracy by more than 60% across all prediction terms. Specifically, the prediction accuracy is approximately 0.05 ns for a 10 min prediction, about 0.16 ns for a 30 min prediction, and remains below 0.3 ns for a 60 min prediction. Therefore, the proposed method provides high-precision real-time LEO satellite clocks in case of long latency of tens of minutes, capable of meeting the demands of decimeter-level real-time PNT services.

During the flight of a space launch vehicle, temporary loss of Global Navigation Satellite System (GNSS) signals can occur due to vehicle dynamics and stage separation events, resulting in navigation solution outages. To address this limitation, integration of a GNSS receiver with an Inertial Navigation System (INS), which provides complementary characteristics, is an effective solution. However, in the current Nuri launch vehicle, which is developed in South Korea, the GNSS receiver and INS operate independently without sharing navigation information, making real-time integrated navigation difficult to implement.

This paper presents the implementation and evaluation of a loosely coupled GNSS/INS integrated navigation algorithm for the Nuri launch vehicle GNSS receiver. The proposed approach utilizes an external commercial-grade INS, the FiberPro FI200P, to provide inertial measurements to the GNSS receiver via RS-422 UART communication. The INS is synchronized using a 1PPS signal generated by the GNSS receiver and delivers inertial data, including accelerations, angular rates, and status information, at a rate of 100 Hz. These measurements are integrated with GNSS navigation data within the receiver using a 16-state Extended Kalman Filter (EKF), which estimates three-dimensional position, velocity, attitude, accelerometer and gyroscope biases, and time delay.

The GNSS receiver is designed to operate under high-dynamic and high-velocity conditions of launch vehicle environments, though its standalone navigation accuracy is relatively low compared to conventional ground or airborne receivers. Vehicle-level experiments were conducted by installing the GNSS receiver, antenna, INS, and a reference GNSS system on a test vehicle and collecting data during urban driving scenarios. The experimental results demonstrate that the integrated navigation solution effectively combines GNSS and inertial information, providing improved continuity and robustness during GNSS outages.

Nevertheless, degradation in navigation performance was observed in certain segments immediately following GNSS outages or during periods of increased GNSS error, due to excessive weighting of GNSS measurements. Future work will focus on tuning the filter parameters (P, Q, and R matrices) and introducing GNSS measurement validation logic to further enhance integrated navigation performance.

The demand for high-accuracy, high-integrity navigation solutions has increased in recent years. Although GNSS carrier-phase positioning techniques such as Precise Point Positioning (PPP) can meet accuracy requirements, no full integrity concept has yet been consolidated or formalized. The PPP technique aims to minimize GNSS measurement errors by applying correction products and models as well as linear combinations of GNSS observables. The remained (residual) errors depend on the correction products quality and the models used, as well as on the signal propagation environment [1, 2]. Potential GNSS faults, such as satellites and/or receiver problems, also need to be considered [3]. Therefore, to design a suitable integrity concept for PPP, its residual errors behaviour needs to be analyzed in different operating scenarios and under nominal and non-nominal / fault-contained conditions. Such an analysis would provide the basis for characterizing nominal errors and developing a PPP integrity concept that can detect and exclude faulty measurements from the navigation solution as well as alert the user in a timely manner.

This study supports the establishment of a PPP-related integrity framework. For this purpose, a state-of-the-art ionosphere-free PPP software with ambiguity resolution (PPP-AR) has been implemented. Ionosphere-free ambiguities are resolved using observable-specific signal biases (OSBs) and applying a two-step procedure: firstly, the wide-lane ambiguities are fixed, followed by the narrow-lane ambiguities fixing. PPP / PPP-AR experiments are conducted for static and kinematic scenarios to draw conclusions for the integrity concept. In the kinematic scenario, we process real flight data that was collected in November 2024 with a Dornier 228 aircraft near Braunschweig-Wolfsburg Airport in Germany. This study examines the resulting PPP and PPP-AR residual errors in order to characterize nominal position errors, and analyses the PPP / PPP-AR performance in terms of accuracy, precision and convergence time. It also assesses the quality of the estimated ambiguities and ambiguity resolution fixing rates. The results are discussed in terms of their implications for establishing a PPP integrity framework, together with open questions and challenges. This work will form the basis for the future PPP integrity concept design.

References:
[1] Xie, M., Wang, N., El-Mowafy, A. et al. (2025) Characterizing PPP ambiguity resolution residuals for precise orbit and clock corrections integrity monitoring. GPS Solut 29, 69. https://doi.org/10.1007/s10291-025-01827-7
[2] Shu, Y., Shi, Y., Xu, P., Niu, X., Liu, J. (2017). Error analysis of high-rate GNSS precise point positioning for seismic wave measurement. Advances in space research, 59(11), 2691-2713.
[3] Du, Y., Wang, J., Rizos, C., El-Mowafy A. (2021) Vulnerabilities and integrity of precise point positioning for intelligent transport systems: overview and analysis. Satell Navig 2, 3 (2021). https://doi.org/10.1186/s43020-020-00034-8