SEBASTIAN CIUBAN, PETER J.G. TEUNISSEN, CHRISTIAN C.J.M. TIBERIUS, DELFT UNIVERSITY OF TECHNOLOGY

Positioning via Global Navigation Satellite Systems (GNSS) and Terrestrial Networked Positioning Systems (TNPS), along with measurements from other sensors (e.g., inertial measurement units, cameras, LiDAR), is widely used or of growing interest in several safety-critical applications, such as automotive, aviation, rail and maritime [1-3]. With these measurements, or observables, a position estimator x̅∈Rⁿ can be formulated, where n≥1 represents the dimension (e.g., n=1 for vertical component, n=2 for horizontal components). Given access to the probability density function (PDF) of x̅∈Rⁿ, denoted f_x̅(x), and an application specific safety-region B⊂Rⁿ (e.g., interval when n=1, ellipse when n=2), one can pose the following question: “What is the probability that the position estimator falls outside the designated safety-region?”

Quantifying this probability enables its comparison against application-specific requirements, or guidelines, to determine whether they are satisfied. This probability can be viewed as a positioning “safety indicator” for the application of interest.

We formulate the aforementioned probability based on the event of positioning failure F=(x̅∉B) [4], and express it as follows

where B^c=Rⁿ\B is the complement of the safety-region (i.e., the Rⁿ space without the safety-region B). The computation of P_F is a challenging task, primarily because the PDF f_x̅(x) may have multiple modes. These modes arise from the position estimator, which, as we assume here, usually results from a combined parameter estimation and statistical hypothesis testing procedure to accommodate for model misspecifications (e.g., faults or outliers in observables). This intricacy is captured in the theoretical framework introduced in [5].

To account for such model misspecifications, one can assume a positioning model that is valid under nominal conditions (the null hypothesis H₀) and design alternative positioning models under multiple alternative hypotheses H_i≠0 with i∈{1,…,k}. For example, in the case of GNSS-based positioning, one can define positioning models under the H_i≠0’s to account for the presence of one or multiple outliers (or faults) in the code-based observables, cycle slips in the phase-based observables, satellite failures, unmodelled atmospheric delays, etc. Then the objective is to select the most likely hypothesis and use the corresponding positioning model to provide the position estimator x̅_i for i∈{0,…,k} (Figure 1). As shown in Figure 1, we highlight that the position estimator x̅∈Rⁿ is a function of the individual estimators x̅_i and the statistical testing procedure used to select these estimators (e.g., a Detection Identification and Adaptation-DIA procedure [6-8] or a Fault Detection and Exclusion-FDE procedure [9-11]).

The expression of f_x̅(x) provides additional insights into its dependencies [5]

where f_x̅i,t̅(x,t) is the joint PDF of the individual estimators x̅_i∈Rⁿ and the vector of misclosures t̅∈R^r used to construct statistical tests for selecting the most likely hypothesis H_i, and p_i(t̅) is an indicator function that equals 1 if H_i is selected and 0 otherwise. The vector of misclosures has the dimension of the measurement redundancy, denoted r, and it contains all the available information useful for testing the validity of the positioning model under H₀. From the expression of f_x̅(x) in (2) we emphasize that P_F from (1) depends on the number of k+1 hypotheses, on the dimension n of the to-be-estimated parameter vector, and also on the dimension r of the misclosure vector. Furthermore, x̅_i∈Rⁿ and t̅∈R^r are dependent for i≠0, which means the joint PDF cannot be expressed as a product of the marginal PDFs (i.e., f_x̅i,t̅(x,t) ≠ f_x̅i(x) f_t̅(t̅), for i≠0). Ignoring or failing to account for the dependence between x̅_i∈Rⁿ and t̅∈R^r, for i≠0, may result in overly-optimistic results when computing P_F, potentially leading to incorrect conclusions that positioning safety requirements are met when they are not [12-15]. The multimodal structure of f_x̅(x) (Figure 2) and its aforementioned dependencies render analytical methods for computing P_F unfeasible.

Another aspect to take into account when computing P_F is related to the requirement for the event of positioning failure F=(x̅∉B)= (x̅∈B^c) to be rare (e.g., P_F<10^-7 [17]). Because analytical methods for computing P_F are unfeasible, numerical integration methods are reasonable candidates. However, as the rarity requirement for F=(x̅∈B^c) becomes more stringent, the computational effort required to compute such probabilities increases significantly. In recent work, we tackled these challenges by introducing a method grounded in rare event simulation principles and proposed a possible approach to address them [18]. The method is intended for use during the design stage of positioning algorithms, where decisions are required regarding (i) measurement models associated with selected positioning technologies and sensors, (ii) parameter estimation techniques for the position vector, (iii) statistical hypothesis testing procedures to accommodate for model misspecifications, and (iv) positioning scenarios of interest (e.g., vehicle driving on a highway, or in an urban area), among other considerations. This methodology is consistent with the scenario-based safety assessment framework widely adopted for studies on automated and autonomous vehicles [19-21].

In this contribution, we outline the principles underlying the approach proposed in our recent work [18]. As an example, we present a simulation-based analysis of positioning safety for an automated vehicle using a real dual-constellation GPS and Galileo satellite geometry [13].

Standard Monte Carlo and Importance Sampling

A reasonable first step is to use numerical integration methods, such as standard Monte Carlo (MC) simulation, to compute P_F. It is possible to re-express (1) as an expected value with respect to the PDF f_x̅(x), denoted E_fx̅(.),

where 1_B^c(x̅) is an indicator function that is 1 if x̅∈B^c, and 0 otherwise. By generating N_x̅ independent and identically distributed (i.i.d.) pseudo-random samples from f_x̅(x) one can approximate (3) as follows (based on counting how many pseudo-random samples fall in B^c=Rⁿ\B)

with its simulation variance, or dispersion, expressed as D(P̅_F^MC)= P_F(1-P_F)/N_x̅. However, the standard Monte Carlo approximation in (4) has limitations when computing probabilities of rare events. Specifically, in the case of rare events a substantial number of pseudo-random samples N_x̅ is needed to be able to compute (4) with a low simulation variance D(P̅_F^MC). For example, if the objective is to compute a target value P_F=10^-9 with √(D(P̅_F^MC))=10^-10, then the required number of pseudo-random samples would be N_x̅≈10¹¹, which involves an excessively large computational effort.

To tackle the limitations of the standard Monte Carlo method, a different approach is needed. Importance Sampling (IS) is a reasonable candidate to be considered as it can achieve simulation variance reduction without a significant increase of the required pseudo-random samples to be generated [22]. The principles of IS have found applicability across a wide area of safety-critical applications, such as safety analyses of structures, nuclear power plants, and for computations of probabilities of collision events in aviation [23-25]. On the basis of IS, we can express P_F from (3), with respect to a newly introduced PDF f̃(x), as follows

where f̃(x) is also called IS density, auxiliar density, or proposal PDF [23, 26]. The main idea is to choose, or find, a proposal PDF that has a larger probability density over the region B^c=Rⁿ\B than f_x̅(x) and 1_B^c(x̅) f̃(x)≠0 whenever 1_B^c(x̅) f_x̅(x)≠0 [27]. If such proposal PDF f̃(x) is chosen, then by generating Ñ_x i.i.d. pseudo-random samples from it allows for the approximation of (5) as follows

The performance of P̅_F^IS in reducing simulation variance depends on the choice of the proposal PDF f̃(x), which in turn drives the quality of the probability estimates. It is possible to find a proposal PDF f̃(x), within a family of PDFs (e.g., exponential family) that minimizes the Kullback-Leibler (KL) divergence, or equivalently the Cross-Entropy, with respect to the optimal theoretical PDF f̃^*(x)=[1_B^c(x̅) f_x̅(x)]/P_F [22]. The theoretical PDF f̃^*(x) is optimal in the sense that it minimizes simulation variance. In practice, f̃^*(x) cannot be used because it depends on the unknown P_F. The proposal PDF f̃(x) that minimizes the KL divergence w.r.t. f̃^*(x) can be determined using the Cross-Entropy method [28].

Probability of Positioning Failure and its Components

We recently proposed a method [18] that can compute the probability of positioning failure from its conditional components (Figure 3). A component-wise computation of P_F enables the determination of the conditional components that contribute most, or least, to its value. Having designed a statistical hypothesis testing procedure to address model misspecifications with a null hypothesis H₀ (comprising of the positioning model believed to be valid under nominal conditions) and k alternative hypotheses H_i≠0 (e.g., comprising of positioning models that account for the presence of outliers, or faults, in the observables), it is possible to decompose P_F into its conditional components based on the statistical testing decisions: Correct Acceptance (CA) when H₀ is accepted and H₀ is valid; False Alarm (FA_i) when H_i≠0 is accepted and H₀ is valid; Missed Detection (MD_i) when H₀ is accepted and H_i≠0 is valid; Correct Identification (CI_i) when H_i≠0 is accepted and H_i≠0 is valid; Wrong Identification (WI_j) when H_j is accepted and H_i is valid for j∉{0,i} (see also the example in Figure 1). This decomposition has been presented and elaborated on in [13]. A graphical representation of the “failure-tree” is shown in Figure 3, where the notation of the probability of positioning failure changed to P_F(b̅) with b̅={b̅₁,…,b̅_k} to account for b̅_i, with i∈{1,…,k}, under all the k alternative hypotheses. Note: Figure 3 goes here The equations that describe the branches and connections in Figure 3 are the following:

where P(H₀) and P(H_i), for i∈{1,…,k}, are the apriori probabilities of the hypotheses H₀ and H_i≠0, P_F|E is the probability of positioning failure conditioned on the statistical testing decision E∈{CA,FA_i,MD_i,CI_i,WI_j}, and P_E is the probability of the event of the statistical decision E. Once the conditional components from (7) are computed, then the total probability of positioning failure is obtained as follows [18],

With (7) and (8) one can obtain the entire characteristic of P_F(b̅) as a function of b̅={b̅₁,…,b̅_k}, thus providing all the information required to perform a rigorous sensitivity analysis for the design purposes of the positioning system. The positioning safety-analysis in the next section is based on constructing the “failure-tree” in Figure 3 via the computations of (7) and (8). Because the size of the outliers in (8) is not known a priori, the maximum value of (8) is reported to provide insights into the worst-case scenario, aiding in the assessment of whether safety requirements or guidelines are met.

Decimeter Level GNSS-Based Positioning Safety Analysis

We consider the GNSS-based positioning safety analysis presented in [13], which is also similar to the one in [18]. The simulation scenario involving an automated vehicle which coordinates are determined in a local East-North-Up (ENU) coordinate system using single-frequency, code-based pseudorange observables in a Differential (DGNSS) setup. The GNSS constellations we consider are GPS (G) and Galileo (E), at L1/E1 frequency. At the considered snapshot of time (epoch), eight GPS and seven Galileo satellites are visible after applying an elevation mask of 10°. Figure 4(a) shows the skyplot of the observed GPS and Galileo satellites by the automated vehicle. Additionally, an elevation depending weighting is applied to the observables.

The horizontal positioning precision is about 0.5 meters (95% circular probability radius).

In the setup of the statistical hypothesis testing procedure to accomodate for outliers or faults in measurements, in this case the DIA procedure, we account for individual outliers, assuming only one occurs at a time (there are k=15 alternative hypotheses). At the output of this procedure the DIA-estimator x̅∈Rⁿ and its PDF f_x̅(x) are obtained. When determining, or choosing, the shape and size of the safety-region B⊂R² several factors should be considered, such as: (i) vehicle’s dimensions, (ii) road geometry to ensure the vehicle is within its lane, (iii) minimum required braking distance as a function of the vehicle’s speed, (iv) proximity with regard to other traffic participants, among other considerations. Several approaches have been proposed in terms of shapes of the safety-region that bound the vehicle (e.g., elliptical, rectangular) in several studies [17, 29, 30]. For our scenario, and for consistency with existing approaches in the literature, we choose an ellipse to inscribe the vehicle, which has a length of 4.5 meters, a width of 1.8 meters, and an orientation of 0° relative to the North axis (Figure 4(b)).

With the DIA-estimator x̅∈Rⁿ, its PDF f_x̅(x), and the ellipsoidal safety-region B⊂R² defined, the method in [18] can be applied for positioning-safety analysis. This yields the computed conditional components of P̅_F(b̅) from equation (8). Figure 5(a) shows the component P̅_F|H0 computed over 50 simulation runs to observe the variability in the results, and Figures 5(b) and 5(c) show the P̅_F|Hi(b̅_i) as a function of the outlier size b̅_i for i∈{1,…,15}. It is noticable that P̅_F|H4(b̅₄), P̅_F|H8(b̅₈), P̅_F|H10(b̅₁₀), and P̅_F|H14(b̅₁₄) are dominating when the size of their respective outlier is larger than 1.60 meters. This can be explained based on the rover receiver-satellite geometry in Figure 5(a), which shows that satellites corresponding to the hypotheses 4, 8, 10, and 14 have a large influence on the horizontal-axis (east-component) of the 2D position solution. Conversely, satellites at low-elevations have a reduced contribution to the 2D position solution (e.g., 6, 9, and 13), which leads to low probabilities of positioning failure under the respective alternative hypotheses.

To compute the maximum P̅_F(b̅), assumptions are needed for the a-priori P(H_i) for i∈{0,…,15}. Because the alternative hypotheses account for outliers in the pseudoranges at the rover-receiver (automated vehicle), it is assumed they primarily occur due to different signal reflections caused by the surrounding environment (e.g., nearby infrastructure). For this analysis, we consider three sets of assumptions ranging from conservative to optimistic cases: (1) P(H₀)=0.98500 and P(H_i)=10^-3 for i∈{1,…,15}; (2) P(H₀)=0.99850 and P(H_i)=10^-4 for i∈{1,…,15}; (3) P(H₀)=0.99985 and P(H_i)=10^-5 for i∈{1,…,15}. The obtained results for the maximum P̅_F(b̅) in the three cases are shown in the Table 1. Note these results correspond to the rover receiver-satellite geometry in Figure 4(a) and the fixed safety-region B⊂R² in Figure 4(b).

In practice, a vehicle will change its orientation while moving (e.g., when making a U-turn, exiting a highway, taking a left/right turn), and consequently, the maximum P̅_F(b̅) will also change. For a short time window (e.g., few minutes), it can be assumed the rover receiver-satellite geometry from Figure 4(a) is constant, allowing us to base our next analysis on the vehicle’s change in the orientation angle θ (measured, in degrees, clockwise with respect to the North axis). Therefore, the safety region B_θ⊂R² (note the change in notation) will also depend on the vehicle’s orientation angle θ. The objective is to compute the maximum P̅_F(b̅) as a function of θ from its components P̅_F|H0 and max_{b̅1,…,b̅15} ∑_i=1¹⁵ P̅_F|Hi(b̅_i). Figure 6(a) shows the results of P̅_F|H0 as a function of θ and that the maximum values is 3.33⋅10^-7±0.0216⋅10^-7 for the vehicle’s orientation angle of θ=110°. In the case of the component max_{b̅1,…,b̅15} ∑_i=1¹⁵ P̅_F|Hi(b̅_i), as shown in Figure 6(b), the maximum value is reached at 1.81⋅10^-2 ±0.00462⋅10^-2 for θ=110°. By combining the results from Figure 6(a) and Figure 6(b) with the assumptions regarding the a-priori probabilities P(H₀) and P(H_i) for i∈{1,…,15} as discussed in the three cases, the results in Figure 6(c) are obtained. Note: Figures 6a, b and c go here In the most conservative case (Case 1), the maximum P̅_F(b̅) at θ=110° is 1.84⋅10^-5 ±0.000216⋅10^-5 while for the most optimistic case (Case 3) the maximum P̅_F(b̅) is 4.94⋅10^-7 ±0.0216⋅10^-7. These results help determine whether the target requirements or guidelines for positioning safety are met. If the requirements or guidelines are not satisfied, it may be necessary to make appropriate changes to the positioning algorithm design choices. These could include aspects such as the measurement setup (e.g., functional and stochastic models), the safety-region, or the combined parameter estimation and statistical hypothesis testing procedure. For instance, the new theoretical framework introduced in [31] shows how fit-for-purpose statistical hypothesis testing improves the performance of DIA-estimators.

Summary and Conclusions

In this contribution, we have presented a new perspective on positioning safety analyses by addressing key ideas and challenges associated with the computation of the probability of positioning failure, such as: (i) the multimodal PDF of the position estimator f_x̅(x), which accounts for the dependence between parameter estimation and statistical hypothesis testing to accommodate potential faults or outliers in the measurement model; (ii) to account for the dependencies on the number of k+1 hypotheses, on the dimension n of the to-be-estimated parameter vector, and also on the dimension r of the misclosure vector when constructing the PDF f_x̅(x) and performing its integration over the region B^c= Rⁿ\B (the Rⁿ space without the safety-region B⊂Rⁿ), which renders analytical methods overly complex or impractical; and (iii) the rarity of positioning failure events in the context of safety-critical applications, which requires more advanced numerical integration methods than standard Monte Carlo.

To address these challenges, we presented an approach based on our recent work in [18], which relies on techniques from rare event simulation, specifically Importance Sampling and the Cross-Entropy Method [22, 28]. The computation and analysis of the probability of positioning failure are intended to be performed during the design stage of positioning algorithms, where key decisions are made regarding (i) measurement models, (ii) parameter estimation methods for the position vector, (iii) statistical hypothesis testing procedures to handle model misspecifications (e.g., outliers or faults in measurements), and (iv) positioning scenarios for vehicles, among other factors. This approach aligns with scenario-based safety assessment frameworks, which are widely used or proposed in studies on automated and autonomous vehicles [19-21].

As an example, we applied the proposed method to perform a single-epoch positioning safety analysis for an automated vehicle, focusing on decimeter-level precision GNSS-based positioning. The method facilitated an analysis of a worst-case scenario aimed at determining the maximum probability of positioning failure. Such analyses can guide decisions on whether positioning safety targets or requirements are satisfied. Once compliance with application-specific requirements is demonstrated based on the probability of positioning failure in the relevant scenarios, the corresponding parameter estimation and statistical hypothesis testing procedure can be implemented for real-time positioning.

While the chosen positioning scenario was centered on the automotive domain, the proposed approach for computing and analyzing the probability of positioning failure is also applicable to other safety-critical fields, including civil aviation, shipping and rail.

Acknowledgements

This research was funded by the Dutch Research Council (NWO) under Grant 18305, titled ‘I-GNSS Positioning for Assisted and Automated Driving.’ The support is gratefully acknowledged.

References 

(1) P. J. G. Teunissen and O. Montenbruck, Eds., Handbook of Global Navigation Satellite Systems. Springer, 2017.

(2) Y. T. J. Morton, F. van Diggelen, J. J. Spilker Jr., B. W. Parkinson, S. Lo, and G. Gao, Eds., Position, Navigation, and Timing Technologies in the 21st Century: Integrated Satellite Navigation, Sensor Systems, and Civil Applications. Wiley, IEEE Press, 2020. 

(3) J. C. J. Koelemeij, et al., “A hybrid optical-wireless network for decimetre-level terrestrial positioning,” Nature, vol. 611, no. 7936, pp. 473-478, Nov. 2022.

(4) RTCA-Special Committee 159, “Minimum Operational Performance Standards (MOPS) for Global Positioning System/Satellite-Based Augmentation System Airborne Equipment,” DO-229F, Radio Technical Commission for Aeronautics, pp. 15, 2020.

(5) P. J. G. Teunissen, “Distributional theory for the DIA method,” Journal of Geodesy, vol. 92, no. 1, pp.59-80, 2018.

(6) W. Baarda, “A Testing Procedure for Use in Geodetic Networks,” Netherlands Geodetic Commission, Publications on Geodesy, 2(5):1–97, 1968.

(7) I. Gillissen and I. A. Elema, “Test results of DIA: A real-time adaptive integrity monitoring procedure, used in an integrated navigation system,“ International Hydrographic Review, vol. 73, nb. 1, pp.75-103, 1996.

(8) P. J. G. Teunissen, “Batch and Recursive Model Validation,” Chapter 24 in Springer Handbook of Global Navigation Satellite Systems, P. J. G. Teunissen and O. Montenbruck, Eds., Springer, pp. 687-720, 2017.

(9) P. T. Hwang and R.G. Brown, “RAIM-FDE Revisited: A New Breakthrough in Availability Performance With nioRAIM (Novel Integrity-Optimized RAIM),” NAVIGATION, 53(1):41–51, 2006.

(10) L. Yang, Y. Li, and C. Rizos, “An enhanced MEMS-INS/GNSS integrated system with fault detection and exclusion capability for land vehicle navigation in urban areas,” GPS Solutions, vol. 18, no. 4, pp.593-603, 2013.

(11) J. Blanch et al., “Baseline Advanced RAIM User Algorithm and Possible Improvements,” IEEE Aerospace and Electronic Systems, 51(1):713–732, 2015.

(12) S. Ciuban, P. J. G. Teunissen, and C. C. J. M. Tiberius, “Dependence Between Parameter Estimation and Statistical Hypothesis Testing: Positioning Safety Analysis for Automated/Autonomous Vehicles,” IEEE Transactions on Intelligent Transporation Systems, vol. 26, no.4, pp. 5509 – 5521, 2025.

(13) S. Ciuban, P. J. G. Teunissen, and C. C. J. M. Tiberius, “GNSS Positioning Safety: Probability of Positioning Failure and its Components,” Proceedings of the 37th International Technical Meeting of the Satellite Division of the Institute of Navigation (ION GNSS+), pp. 2228-2249, 2024.

(14) S. Zaminpardaz and P. J. G. Teunissen, “On the computation of confidence regions and error ellipses: A critical appraisal,” Journal of Geodesy, vol. 96, no. 10, pp.1-18, 2022.

(15) S. Zaminpardaz, P. J. G. Teunissen, and C. C. J. M. Tiberius, “Risking to underestimate the integrity risk,” GPS Solutions, 23(29):1–16, 2019.

(16) Wikimedia Commons, Citroen C3, top. Available: https://commons.wikimedia.org/wiki/File:C3top.png#file. 

(17) Reid, T. G. R. et al., “Localization Requirements for Autonomous Vehicles,” in SAE International Journal of Connected and Automated Vehicles, 2019.

(18) S. Ciuban, P. J. G. Teunissen, and C. C. J. M. Tiberius, “A Method to Compute the Probability of Positioning Failure for Vehicles in the Context of Dependence Between Parameter Estimation and Statistical Hypothesis Testing,” IEEE Transactions on Vehicular Technologies, vol. 74, no.10, pp. 15238 – 15253, 2025.

(19) S. Riedmaier et al., “Survey on Scenario-Based Safety Assessment of Automated Vehicles,” IEEE Access, vol. 8, pp.87456-87477, 2020.

(20) U.N.E.C.E., “New Assessment/Test Method for Automated Driving (NATM) Guidelines for Validating Automated Driving Systems (ADS),” United Nations Economic Commission for Europe–Inland Transport Committee, Report, 2023.

(21) E. de Gelder et al., “TNO Street Wise: Scenario-Based Safety Assessment of Automated Driving Systems,” Netherlands Organisation for Applied Scientific Research (TNO), White Paper, 2024.

(22) H. Kahn and A. W. Marshall, “Methods of Reducing Sample Size in Monte Carlo Computations,” Journal of the Operations Research Society of America, vol. 1, no. 5, pp.263-278, 1953.

(23) I. Papaioannou, C. Papadimitriou, and D. Straub, “Sequential Importance Sampling for Structural Reliability Analysis,” Structural Safety, vol. 62, pp.~66–75, 2016.

(24) B. J. Garrick, et al., Reliability analysis of nuclear power plant protective systems (Research and Development Report). Holmes and Narver Inc. Nuclear Division, 1967.

(25) M. Mitici and H. A. P. Blom, “Mathematical Models for Air Traffic Conflict and Collision Probability Estimation,” IEEE Transaction on Intelligent Transporation Systems, vol. 20, no. 3, pp.1052-1068, 2019.

(26) R. V. Rubinstein and D. P. Kroese, Simulation and the Monte Carlo Method. Wiley Series in Probability and Statistics, 2008.

(27) G. Biondini, “An introduction to rare event simulation and importance sampling,” Chapter 2 in Handbook of Statistics, V. Govindaraju, V. V. Raghavan, and C. R. Rao, Eds., Elsevier B.V., pp. 29-68, 2015. 

(28) R. V. Rubinstein and D. P. Kroese, The Cross-Entropy Method: A Unified Approach to Combinatorial Optimization, Monte-Carlo Simulation, and Machine Learning. Springer Series in Information Science and Statistics, 2004.

(29) Y. Feng, C. Wang, and C. Karl, “Determination of Required Positioning Integrity Parameters for Design of Vehicle Safety Applications,” ION GNSS+, pp. 129-141, 2018.

(30) O. N. Kigotho and J. H. Rife, “Comparison of Rectangular and Elliptical Alert Limits for Lane-Keeping Applications,” ION GNSS+, pp. 93-104, 2021.

(31) P. J. G. Teunissen, “On the Optimality of DIA-Estimators: Theory and Applications,” Journal of Geodesy, vol. 98, no. 43, pp.1-26, 2024.

Authors

Sebastian Ciuban received the M.Sc. in Aerospace Systems: Navigation and Telecommunications from École Nationale de l’Aviation Civile (ÉNAC), Toulouse, France, in 2017. Post-graduation, he joined the European Space Research and Technology Centre (ESTEC) as a Young Graduate Trainee (YGT) in the Directorate of Navigation. From 2019 to 2021, he worked as a GNSS Engineer at CGI Nederland B.V. In 2021, he started a Ph.D at Delft University of Technology, Delft, The Netherlands, in the field of PNT safety for automated and autonomous vehicles (to be defended in December 2025). As of June 2025, he joined the Defense and Security Unit of Science and Technology (S[&]T) B.V, Delft, The Netherlands.

Peter J.G. Teunissen is Professor of Geodesy and Satellite Navigation at Delft University of Technology, the Netherlands, and a member of the Royal Netherlands Academy of Arts and Sciences. His past academic positions include Head of the Delft Earth Observation Institute, Science Director of the Australian Centre for Spatial Information, and Federation Fellow of the Australian Research Council. He has been research-active in various fields of Earth Observation, with current research focused on the development of theory, models and algorithms for high-accuracy applications of satellite navigation and remote sensing systems.

Christian C.J.M. Tiberius received a Ph.D. on recursive data processing for kinematic GPS surveying from Delft University of Technology, Delft, The Netherlands. He is an Associate Professor at the Geoscience and Remote Sensing (GRS) Department, Delft University of Technology. His research interests include navigation with GNSS and terrestrial radio positioning, primarily for automotive applications.

The post Positioning Safety for Safety-Critical Applications via Probability of Positioning Failure appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

The project is a direct response to Czech railway operators calling for a modular, high-accuracy GNSS-based device, with Galileo as the preferred satellite system. But the resulting solution is easily adaptable to other sectors requiring precise positioning and reliability.

NavCom X combines advanced hardware and a distributed services software architecture, enabling precise, standalone tracking and system monitoring, independent of external power or connectivity. The system is designed for high performance, resilience, and flexibility, capable of tracking service intervals and accessing real-time and historical data.

The system also features integrated Galileo OSNMA (Open Service Navigation Message Authentication), providing authentication of navigation messages to mitigate spoofing. NavCom X anticipates future ERA (EU Railways Agency) standards, including those pertaining to OSNMA.

Hardware and software integration

At a recent ESA-hosted event, NavCom X team leaders, including Petr Seč from Betrian and Tomáš Duša from GCE, presented the system. The hardware design is compact and robust. Described as ‘not just a bulky tablet’, it supports external antennas, strong power supply options and a user interface tailored to specific use cases.

Energy efficiency was a major project focus. The software architecture includes modular daemons (background processes), such as GNSS processing, OSNMA handling and data transmission. The team followed receiver guidelines and conducted early-stage tests using software-defined radios (SDRs) and vectors provided by the European Agency for the Space Program (EUSPA).

A comprehensive testing campaign was conducted at the EU Joint Research Centre (JRC), including cold, warm, and hot start scenarios, as well as spoofing resilience trials using hardware-in-the-loop (HIL) techniques and GNSS E1 signal spoofing.

Field trials took place at the Velim test rail circuit in the Czech Republic, operated by the Railway Research Institute (VUZ). The campaign covered 20 test scenarios under realistic conditions, involving multiple locomotive types, varying speeds, and with L1 jamming. Mechanical, electrical, UI, connectivity, and data handling were all verified. All navigation performance goals were met, OSNMA authentication and resilience against spoofing were demonstrated successfully, with battery performance exceeding targets.

Seč said the NavCom X system sets a new foundation for safe, efficient, and scalable GNSS adoption in rail and other sectors such as agriculture, high-definition mapping, and construction. Next steps include improving modularity, refining OSNMA features, enhancing energy performance, and integrating cloud-based analytics.

The post Compact and Flexible NavCom X for Rail and Beyond appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

Andreas Wenz of SBB, along with Michael Roth of the German Space Agency (DLR) and Paulo Mendes of Industrieanlagen-Betriebsgesellschaft (IABG), presented the results of the project at a special event hosted by the European Space Agency (ESA).

The long-term goal, Wenz explained, is to introduce new positioning technologies such as GNSS, into the control command and signaling scheme of the European Rail Traffic Management System (ERTMS). This will enable so-called ‘moving block’ operations, where the safe distance between trains is calculated in real-time based on their actual speed and braking capabilities.

Under the conventional ‘fixed block’ system, a railway line is divided into predefined, static blocks or sections. Only one train can occupy a block at any time, regardless of its speed or braking ability, thus limiting capacity and efficiency. Under the moving block approach, supported by ERTMS Level 3, a safe zone or ‘protection zone’ is continuously calculated behind each train. This zone moves with the train, hence ‘moving block’, allowing other trains to follow much more closely. For this to work, operators need accurate, real-time data about train position, speed, and braking ability.

Moving forward

ERTMS moving block operations rely on continuous communication, via GSM-R or newer systems, and onboard train integrity monitoring. Benefits include increased line capacity, shorter time between trains, more efficient traffic flow, reduced need for trackside assets, and lower costs.

EGNSS MATE algorithm development, undertaken by DLR, was accompanied by a measurement campaign using SBB vehicles on the Swiss normal gauge network. Collected sensor data was used to test the algorithms. IABG analyzed the use of new Galileo services OSNMA and HAS within the ERTMS, and tested the performance of the new algorithms against jamming and spoofing attacks, as simulated in the laboratory.

Wenz said the results of the project will help European rail operators to standardize localization solutions and will enable new product development within the signaling industry. The project plans to allow full access to the developed algorithms and will create a test catalogue to serve as a basis for future product certification.

The EGNSS MATE project was co-funded under ESA’s NAVISP program, Element 2, which aims to increase the competitiveness of the European PNT industry.

The post ESA-funded Project Enables ‘Moving Block’ Operations in European Rail appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

Speaking from Intecs headquarters in Italy, Patrizia Nencioni and Stefano Esposito Mocerino explained the problem and the solution. “Knowing the position of the train is of vital importance for correct and safe traffic control,” Nencioni said. To track train positions, the current European ERTMS/ETCS signalling system employs balises. These are electronic beacons or transponders placed between the rails at known positions. A train that loses its position, for example when the on-board system has to be restarted, has to move at low speed until it picks up the next balise signal. Once it has thus re-established its position, it can then accelerate to its standard operating speed.

Because railway belises can be separated by some kilometers, the delay when a train loses its position can be significant, affecting other vehicles on the rail lines. The Intecs system, dubbed AGIS4RAIL (‘assisted GNSS with imaging sensors for rail applications), enables determination of position without waiting for the train to reach the next balise, reducing the time required for it to gain full speed.

Test campaign

A number of field tests of the new system have been undertaken. A crucial set of trials, carried out at an auto racing track, involved a ground vehicle with GNSS antennas on the roof and with cameras pointing to one side, where QR code panels were set up at 10-meter intervals. The vehicle completed 30 laps around the course under different conditions. AGIS4RAIL correctly identified the position of the vehicle at every lap, with the vehicle attaining a maximum speed of 35 km/h, which is 5 km/h above the operating speed of the type of train that would use the new system. The maximum estimated error was 4.76 m, in line with the target of 5 m.

Mocerino said, “We believe our system is capable of delivering improved positioning performance both in terms of availability and integrity.” The use of imaging sensors mitigates all of those well known local effects that can degrade the performance of GNSS-only-based positioning systems, such as multipath. AGIS4RAIL also provides resiliency against jamming and spoofing. The QR-code landmarks can include data for the authentication of the landmark itself. The low cost of a QR code panel when compared to an electronic belise is also to be considered.

Intecs is now actively seeking an industrial partner with whom it will work towards product commercialization. In the meantime, further testing goes on, soon to include trials on-board operational trains in Italy. The AGIS4RAIL project received funding under the European Space Agency’s NAVISP program.

The post Intecs Fusing GNSS and Other Sensor Data for Train Localization appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

After the successful production launch in 2020 of the ±14 g AXO315 accelerometer for high-performance navigation and positioning of dynamic systems, Tronics extends the AXO300 accelerometer series with AXO301, a low-noise and high-resolution ±1 g accelerometer for high precision acceleration/deceleration measurements in railway applications and inclination control in industrial applications, and AXO305, a ±5 g accelerometer tailored for navigation, positioning, and motion control of land and marine manned and unmanned systems.

Built with an innovative closed-loop architecture that delivers high linearity and stability even under strong vibrations, the accelerometers from the AXO300 platform feature an excellent one-year composite bias repeatability of 1 mg and composite scale factor repeatability of 600 ppm.

AXO301 High-Resolution Accelerometer and Inclinometer for Railway and Industrial Systems
AXO301 is a low-noise, high-resolution, closed-loop digital MEMS accelerometer with ±1 g input range that offers a performance-equivalent, low-SWaP (size, weight and power) and cost-effective alternative to force balance inclinometers and servo-accelerometers. It demonstrates an ultra-low noise density of 8 μg/√Hz with an excellent 50 μg resolution to offer high-accuracy inclination angle measurements. AXO301 is tailored to odometry assistance for train positioning and localization systems, high-end industrial tilt and inclination measurements systems as well as motion control of construction machinery. The AXO301 is compliant with EN61373 railway standard for vibrations and shocks.

AXO305 High-Performance Accelerometer for Land, Marine and Robotics applications
With an input measurement range of ±5 g and vibration rectification error of 20 μg/g², AXO305 is tailored to navigation, positioning and motion control functions of land, rail and marine transportation systems and vehicles.

It demonstrates a Bias Instability of 4 μg with a ±0.5 mg bias over its temperature range, thus enabling precise GNSS-aided navigation of manned and unmanned ground vehicles and trains when integrated into inertial navigation systems. AXO305 is a perfect candidate for motion reference units used for ship motion control and dynamic positioning, inertial measurement units for land navigation, subsea navigation of autonomous underwater vehicles and remotely operated vehicles, platform and crane stabilization as well as precision robotics.

The closed-loop architecture of Tronics AXO300 platform offers high resolution and strong vibration rejection. Accelerometers and inclinometers from the Tronics AXO300 series are housed in a miniature, hermetic, ceramic J-lead package that ensures long operational and storage life and guarantees a high compliancy with the stringent thermal cycling requirements of critical applications. They embed fully hard-coded electronics with a 24-bit digital SPI interface for a swift integration into stand-alone sensor modules, INS, IMU as well as attitude and heading reference systems. The built-in self-test ensures initial verification of the sensor’s integrity and continuous in-operation functionality test.

Low SWaP

Thanks to their common sensor’s architecture, miniature package and low-power consumption, Tronics AXO315, AXO305 and AXO301 accelerometers offer a digital, cost-effective and low-SwaP alternative to bulky, expensive, and power-consuming analog solutions like tactical-grade quartz accelerometers. AXO300 accelerometers are ideally complemented by high performance Tronics GYPRO digital rate gyros that share the same SMD J-lead ceramic package (12 x 12 x 5 mm) and same digital interface to enable low-cost integration, assembly, and reliability on PCB, even in fast-changing temperature conditions.

AXO315 volume production started in 2020. AXO301 and AXO305 are now available for sampling and customer evaluations, directly at Tronics or through specialized distribution channels like Texim. Swift evaluation of the sensors can also be made with an Arduino-based evaluation kit that provides built-in testing functionalities such as output reading and recording, recalibration, and digital self-tests.

The post TDK Expands Tronics AXO300 Series with Two Types of Digital MEMS Accelerometer Sensors appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

In 2021, GNSS and EO downstream market generated over 200 billion euros revenues and are set to reach almost half a trillion over the next decade. EO and GNSS data have become increasingly important in the big data realm and paradigm responding to natural and man-made disasters, curbing the spread of disease and strengthening a global supply chain, among many other goals.

The Report provides analytical information on the dynamic GNSS and EO markets, along with in-depth analyses of the latest global trends and developments through illustrated examples and use cases. Using advanced econometric models, it also offers market evolution forecasts of GNSS shipments or EO revenues spanning to 2031.

With a focus on Galileo/EGNOS and Copernicus, the report highlights the essential role of space data across 17 market segments including,

• Agriculture; Aviation and Drones;
• Biodiversity, Ecosystems and Natural Capital;
• Climate Services; Consumer Solutions, Tourism, and Health;
• Emergency Management and Humanitarian Aid;
• Energy and Raw Materials; Environmental Monitoring;
• Fisheries and Aquaculture; Forestry;
• Infrastructure;
• Insurance and Finance;
• Maritime and Inland Waterways;
• Rail;
• • Road and Automotive;
• Urban Development and Cultural Heritage;
• and Space.

Some report highlights:

• Global annual GNSS receiver shipments will reach 2.5 bn units by 2031, dominated by the applications falling under the Consumer Solutions, Tourism and Health segment contributing roughly to 92% of global annual shipments;

• In EO, aside from the largest markets like Agriculture, Urban Development and Cultural Heritage or Energy and Raw Materials, the Insurance and Finance segment is expected to experience the fastest growth over the next decade (21 % of CAGR) for both EO data and value-added service revenues;

• From a supply perspective, the European Industry holds over 41% of the global EO downstream market and 25 % of the global GNSS downstream market.

“The flagship EU Space Programme, driven by Galileo and EGNOS on one side and Copernicus on the other, has become a major enabler in the downstream space application market. As a user-oriented agency, we provide this inside information on markets evolution to our users, being innovators, entrepreneurs, investors, academic researchers, chipset manufacturers, or simply anyone who looks into space to bring value to their activities. The added value and key differentiators of European GNSS and EO are showcased, both separately and in synergy with each other. I know that the report will be of great use and inspiration for those who are contributing to the EU economic growth,” concluded EUSPA Executive Director Rodrigo da Costa.

The report is available for download here.

The post GNSS and Earth Observation Market Report Finds 200 Billion Euro ($229 Billion) Revenue Generated in 2021 appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

The online interactive map is accessible here.

Augmentation service providers deliver a range of high-accuracy GNSS positioning services worldwide, tailored both for professional and consumer markets. Service providers monitor signals from GNSS (GPS, GLONASS, Beidou and Galileo) satellites and generate corrections to significantly improve the accuracy of GNSS standalone positioning. Correction messages are transmitted via the internet, SATCOM or GPRS to GNSS receivers. There are different types of services appropriate to all needs and budgets, offering different levels of accuracy from centimeters to decimeters. These solutions can be used in a number of markets including: mapping, surveying, construction, agriculture, automotive or aviation, to name a few.

Standard GNSS positioning, often affected by several errors, can be corrected using augmentation services to provide a more accurate and precise position. With augmentation services, users can operate their receivers virtually anywhere on the globe and, by means of receiving data from a control center, achieve accuracies ranging from meter to centimeter-level, depending on the hardware, platform and application.

Today, there are different augmentation techniques based on the use of a network of ground-based reference or monitoring stations with known locations that enable to calculate corrections (e.g. differential corrections for RTK or clocks and orbits corrections for PPP). These corrections can then be disseminated, for instance, over the internet or satellites.

The augmentation service provider map is a thematic map updated on a quarterly basis that provides an easy way to visualize the service providers that are able to operate in each country. Clicking on each country shows the names of the Galileo-ready providers along with the name and type of service and coverage.

The post Map of Augmentation Service Providers Supporting Galileo Now Available appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

“We are currently developing and testing with success SBAS next-generation architectures and capabilities. The strong dynamic we experience on our SBAS export markets in Asia, Africa & Indian Ocean demonstrate our solution global attractiveness for our customers,” said Benoit Broudy, Navigation Vice President at Thales Alenia Space in France.

Thales Alenia Space is a joint venture between the French company Thales (67%) and the Italian company Leonardo (33%).

The post EGNOS Contract for Next-Generation, Dual-Frequency European SBAS Signed appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

The European Railway Traffic Management System (ERTMS) is a major industrial project implemented by the EU to create an interoperable railway system in Europe that is safer and more efficient.

The proposed solution is based on multi-sensor fusion using measurements from a GNSS receiver, an inertial measurement unit (IMU) and a tachometer with the support of a digital map of the rail tracks. The localization system consists of a data fusion algorithm associated with an integrity algorithm, ensuring the SIL4 level of safety of the main outputs of the Train Localisation On Board Unit (TLOBU). The integrity algorithm uses the European Geostationary Navigation Overlay Service (EGNOS,) a satellite-based augmentation system.

Originally specified by aviation users, EGNOS provides safe augmentation information for GNSS, composed of navigation data corrections (orbits and time) and of integrity information. EGNOS V2, currently operational, augments only the American GPS.

EGNOS V3, currently under development by Airbus Defence and Space, will provide improved performances for aviation users, and in addition will augment both American GPS and  European Galileo with its upcoming DFMC (Dual Frequency Multi-Constellation) release.

Whys and Wherefores

The TLOBU will provide trains and railway operators with critical information such as positioning and velocity, complemented by acceleration, heading and attitude for non-safe applications.

Rail is one of the most environmentally friendly modes of transport. In the European Union (EU), rail is responsible for less than 0.5% of transport-related greenhouse gas emissions. This makes it one of the most sustainable forms of passenger and freight transport.

Knowing the exact position of a train is at the heart of rail operations across the EU. It aids rail operators in efficient train traffic management and also informs passengers, both onboard and waiting at the station, to know whether their train is delayed.

To ensure EU-wide interoperability, real-time, precise train positioning and high levels of safety, the ERTMS currently relies on a series of costly ground instruments. In the coming years, this will change, and ERTMS may switch to EU space solutions. CLUG, sponsored by the European Union Agency for the Space Program (EUSPA) has made concrete steps towards providing a cost-efficient train tracking solution based on EU satellite technology together with other sensors and data.

The project’s goal is to assess the creation of a failsafe TLOBU that will be interoperable across the entire European railway network.  Based on experience gained during the demonstration phase, the consortium will collect and review data that will help rail operators and industry to gain insights and push towards a new version of the ERTMS standards.

Data from the TLOBU are transmitted to specific train safety functions such as the European Vital Computer (EVC), part of the Automatic Train Protection function (ATP). In parallel, the fusion algorithm is also providing other outputs to other train functions that do not require a SIL4 level of safety, such as Train Management System (TMS) or the passenger information system.

Architecture and Algorithms

The system architecture and algorithms are defined by Airbus Defence & Space, and NAVENTIK. Both companies are providing two different solutions for the fusion algorithms, whereas the integrity concept is defined by Airbus Defence & Space. This concept is based on the EGNOS services; however, the currently available services have only been defined for aviation means and requires specific refinements to be optimized for rail environments. This EGNOS service is the cornerstone of the integrity concept of CLUG to reach the necessary SIL4 level of safety. Airbus D&S detailed this EGNOS service for rail in specific deliverables, which will be published in the coming months.

Siemens and its partners are performing data collection on three different trains in Switzerland, France and Germany. All this data will then be used to test the localisation algorithms designed in the framework of the project in order to gather as much as possible experience and knowledge about the behaviour of these sensors and their associated fusion algorithms in railway environment.

From the start of the data collection in November 2020 until the end of May 2021, data has already been collected over nearly 2,000 hours and 45,000 km, a distance of more than once around the world.

The shared raw data will then be processed in two separate test environments, developed by Siemens and Naventik, using sensor fusion algorithms developed by Airbus and Naventik. This step emulates possible positioning solutions based on the two types of fusion algorithms, generating position and velocity information as if these systems would have been installed onboard the train during the raw data collection. Performing this sensor fusion offline however makes it possible to “re-run” the same journey again and again, using improved versions of fusion algorithms.

TLOBUs to Replace Balises

Using EU space technology in the railway sector not only increases safety but can significantly reduce maintenance and other operational costs. This new approach for train localization is set to improve the current system based on balise readers. A balise is an electronic beacon or transponder placed between the rails of a railway as part of an automatic train protection (ATP) system.

The goal of the TLOBU is to ultimately replace the current localization system, and thus to promote and accelerate the deployment of ERTMS in Europe by introducing more accurate train localization. Such an innovative system should also help drastically reduce the ground equipment, currently ensuring the safe train localization, such as axel counters and track circuits. Although one of the goals is to decrease as much as possible the use of balises along the tracks, the system will still make use of some balises to help maintain a precise and safe position in GNSS-denied environments such as tunnels and train stations.

The post EGNOS a Safe, Efficient Locator for Europe’s Trains appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

A free webinar on August 31 explains how an inertial measurement unit (IMU) assembles and delivers precise timing information, typically at a much higher rate than the co-integrated GNSS receiver.

[Photo above: OWLS from Latronix AB will be discussed, dissected and analyzed during the webinar. For contact-free measurement of track, rails and overhead lines, OWLS delivers high-resolution data at speeds up to 300 km/h. This enables the installation of the equipment on vehicles in regular traffic. The OWLS device integrates speed sensors, tachometer, GPS, lasers, optical sensors and inertial measurement components. Timing is everything!]

Attendees will learn how correct synchronization of this data across the system is critical to performance in the application, particularly in high-dynamic conditions, challenging environments and autonomy.

Register here for the free August 31 webinar, “The Right Time for the Right Place.”

An acknowledged expert in the field, with more than 30 years of cutting-edge experience in multi-sensor R&D, opens the discussion with an examination of the internal workings of the IMU and how the user can access timing information within it for ultimate advantage. A specialist in IMU design and fabrication then explores the advantages furnished by the latest micro-electromechanical systems (MEMS) technology and how the customer can best utilize options furnished within the IMU.

Finally, an engineer focused on industrial inspections examines the IMUs role in successful application, discussing the critical aspect of synchronization of the different sensors within an IMU – GNSS integration. Challenges and solutions related to timing and design integration of multiple sensor inputs for pre- and post-analysis will be covered.

What you’ll gain from this webinar:

• An understanding of IMU architecture
• Knowledge of the timing construction process inside the IMU and how it is synchronized across the system
• Inside access to this critical timing data
• Special considerations with respect to timing in the IMU
• Instruction on use of various options within the IMU
• An appreciation of timing error impact on application performance

Our expert panelists:

John Raquet, Director at Integrated Solutions for Systems (IS4S)-Dayton, where develops efficient, pluggable approaches to all-source navigation. He founded and was formerly the Director of the Autonomy and Navigation Technology (ANT) Center at the Air Force Institute of Technology (AFIT). He has a multidisciplinary background-teaching in electrical engineering with a PhD in geomatics engineering from the University of Calgary, a masters in aero/astro engineering from the Massachusetts Institute of Technology, and a BS in astronautical engineering from the US Air Force Academy. He is a past President of the Institute of Navigation (ION), has been a US Fulbright Scholar, and is an ION Fellow. He has been developing navigation system technology for more than 30 years.

Reidar Holm is a Product Development Manager at Sensonor, a producer and developer of high-precision, light-weight gyros and IMUs. He works MEMS R&D and design, ASIC design, low-stress package design, system design, assembly and calibration, and high-volume production for automotive, MEMS pressure sensors, accelerometers, gyros and IMUs. He has a degree in electrical engineering and electronics from the University of Manchester Institute for Science and Technology

Björn Skatt, Chief Technology Officer and R&D Coordinator at Latronix AB, a Swedish railway inspection firm, where he specializes in system design, image processing, geometrical calculations, 3D visualizations and hardware-connected programming for optical and inertial measuring systems for railway applications.

Register here for the free August 31 webinar, “The Right Time for the Right Place.”

The post For PNT Integration, Timing Is Everything appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.