A close look at several options that identify potential improvements at system level for a reasonable increase in complexity.

ÉLINE RENAZÉ, CHRISTOPHE BOURGA, MATTHIEU CLERGEAUD, THALES ALENIA SPACE

JARON SAMSON, EUROPEAN SPACE AGENCY (ESA)

The Satellite-Based Augmentation System (SBAS) integrity concept expects the system to alert the user of “out of tolerance conditions” within a required delay that’s compatible with intended operations. This delay is called the time to alert (TTA).

The SBAS TTA is the part of time to alert allocated to SBAS. It corresponds to the maximum allowed time elapsed between the “out of tolerance condition” and the reception of an alarm at the user level. When using precise differential corrections, an out-of-tolerance condition is defined as a horizontal error exceeding the Horizontal Protection Level (HPLSBAS) or a vertical error exceeding the Vertical Protection Level (VPLSBAS) [1] attachment B 3.5.7.5.1.

The European Geostationary Navigation Overlay Service (EGNOS) currently complies with the 6-second time to alert requirement compatible with Category I precision approach operations as defined in table 3.7.2.4.2-1 [1] (split into 5.2s for the SBAS system and 0.8s for the user display).

Early feedback from users indicates there may be an interest for future services offering a TTA budget smaller than the TTA budget of SBAS systems. The target for global TTA is between 1.5 and 5 seconds (allowing to achieve up to Category II approach TTA minima).

To prepare this evolution for further EGNOS versions, Thales Alenia Space has studied several evolutions to reduce the SBAS TTA. This study is system driven, identifying potential improvements at system level, for a reasonable increase of complexity. 

A SBAS computes corrections and integrity data based on GNSS observations that are performed by the ground reference station network (named Receiver Integrity Monitoring Stations (RIMS) in EGNOS). In the context of a Dual Frequency Multi Constellation (DFMC) service, the observations contain the GNSS and SBAS navigation messages received and carrier-code, carrier-phase measurements on L1, L5 (for GPS) and E1, E5a (for Galileo) frequencies. 

The SBAS also uses these observations to detect out of tolerance conditions. 

The corrections, integrity data and alerts are then sent through SBAS messages using a GEO satellite. Observations are computed at RIMS level at 1Hz frequency, which corresponds to the frequency at which SBAS messages are broadcast.

Figure 1 provides a first level decomposition of the time to alert.

The “out of tolerance condition” is detectable from the SBAS using the next observations done at ground station level. Knowing the observations are performed at 1Hz frequency, the “waiting for next observation” time is between 0 and 1,000 ms (1,000 ms in worst case).

The SBAS processing time corresponds to the time needed by the system to process the observations, generate the corresponding NOF (Navigation Overlay Frame) message and prepare it for broadcasting.

The NOF message is then sent at the next GPS second. This generates a “waiting for next broadcasting in NOF” time. This time is between 0 and 1,000 ms.

The SBAS processing time is spread over the SBAS systems. Figure 2 provides a high-level overview of the EGNOS V2 system.

The system is composed of the following subsystems:

• The Receiver Integrity Monitoring Stations (RIMS) that collect GNSS/GEO navigation messages and provide GNSS measurements to the Central Processing Facility (CPF). 

• The CPF computes the complete navigation context, provides corrections and integrity data based on the data provided by RIMS and generates the NOF message to be broadcast

• The Navigation Land Earth Stations (NLES) that broadcast the NOF message to the GEO satellite

• The GEO satellite that broadcasts the NOF Signal In Space (SIS) to users

• The EGNOS Wide Area Network (EWAN) that is responsible for the data transmission from one subsystem to another

Each of the previously defined subsystems contribute to the SBAS processing time included in the time to alert.

Figure 3 gives the EGNOS V2 TTA budget allocation.

This article presents the optimizations axis for each part of TTA, except the user processing part that cannot be improved by the SBAS.

Optimization of SBAS Processing Time

The SBAS processing time is the part of TTA that is completely under SBAS responsibility.

It is composed of observations processing at RIMS level, transmission of observations from RIMS to CPF through EWAN, CPF processing, transmission of NOF message from CPF to NLES through EWAN and NLES processing.

Figure 4 provides a high-level decomposition of SBAS processing time based on EGNOS V2 allocation.

An important part of the budget allocated to SBAS processing is dedicated to transit delay through EWAN (in orange in the previous graph).

The transit delays through EWAN constitute an important part of the TTA budget allocated to SBAS processing. 

The collocation of CPF and NLES makes it possible to avoid transit delay between CPF and NLES. Based on EGNOS V2 allocation, this reduces SBAS processing time to 200 ms.

The transit delay between RIMS and CPF is mainly driven by the technology used for the link.

There are two types of RIMS sites:

• Wired RIMS. RIMS connected to a terrestrial access line.

• VSAT RIMS. RIMS connected to a Hub VSAT (a teleport) via a satellite link.

VSAT links are used on less than 25% of the RIMS sites on EGNOS V2.

Even if the RIMS to CPF maximum transit delay requirement is the same for both types, the observed transit delay is higher for VSAT than for Wired RIMS as shown in the next analysis.

Observation of transit delay on EGNOS V2 over a 11-day period gives the following results for all available VSAT RIMS and wired RIMS:

A solution without VSAT links reduces the allocated time to about 500 ms without the need for further technology improvement.

750 ms are allocated to CPF processing. The CPF is composed of two components:

• Processing Set (PS) that ensures the correction and integrity data and constructs the navigation message.

• Check Set (CS) that ensures the system integrity and verifies the generated messages.

Both components shall be analyzed. The CPF execution time is driven by the maximum execution time of both components and by the exchanges between PS and CS and with other subsystems like NLES. The exchanges are taken into account in the analysis.

Two axes of improvements have been studied: An update of hardware and algorithms with their scheduling improvements.

A previous analysis of PS algorithms (legacy and new algorithms) response time was performed using EGNOS V2 implementation, keeping the same algorithm sequencing and repartition over boards on a Linux Platform with a 3.5GHz CPU. Following this experimentation, 32 ms are theoretically needed per cycle to perform the computation of the latest V2 algorithms. 

This leads us to consider an overall estimation of about 75 ms of total execution per cycle with a 1.5GHz CPU on nominal conditions (in single frequency context). Considering Dual Frequency Multi Constellation context (about 25% of additional computation time), about 25% of overhead for management of external interface (CS, NLES) and inter algorithm management, and an additional margin on about 30%, about 150 ms to 200 ms is needed for EGNOS V2 algorithms processing within the CPF allocated time. 

Requirements for available CPU margins also have been considered. The analysis shows considerable margins exist in terms of CPU load to achieve a reduction of TTA budget for the Processing Set. 

The analysis concludes a hardware composed of six boards with 1.5GHz CPU, reducing the CPF Procession Set execution time to 200 ms.

An equivalent analysis was performed with the CPF Check Set. The analysis was performed using EGNOS V2 implementation on a multi-board architecture based onVM6052 running at 2.2GHz.

This analysis concludes the processing time can be reduced to 200/300 ms. 

In the context of DFMC, as part of the algorithms (inter frequency bias) are not needed, we can conclude that CPF processing time can be easily reduced to 250 ms.

In addition to EGNOS V2, Thales Alenia Space has developed its own Processing Set with a single board architecture (VM6052 running at 2.2GHz). By using SW partitioning, all algorithms can be scheduled on a single board with CPU margins compliant with resource consumption requirements (CPU is only used at 30% with a 2.2GHz processor). 

Taking into account the CPU consumption on a single frequency implementation, and adaptation to DFMC (expandability factors for new constellation and non-required algorithms), the algorithms can be processed in less than 200 ms.

Taking into account these analysis, CPF processing can be reduced to 250 ms (4Hz).

Taking into account optimization of transit delays and CPF processing time, the SBAS processing time can be reduced to 905 ms as shown in Figure 5.

Optimization of waiting for next measurement time

As shown in Figure 1, part of the delay to alarm is due to the time elapsed between the “out of tolerance” event and the observation at RIMS level.

Several axes have been analyzed to reduce this time:

• Introduction of asynchronous processing. This makes it possible to detect “out of tolerance” events as soon as enough observations are available.

• Increase of RIMS output rate. Observations are performed more often at RIMS level, reducing the time between the “out of tolerance” event and the next observation.

Asynchronous Processing

In current EGNOS V2 implementation, observations are performed at the same time on all RIMS synchronized with GPS time. CPF algorithms are executed at 1Hz frequency as soon as almost all RIMS measurements are received.

In an asynchronous scheme, the RIMS produces measurements not synchronized with GPS time. The generation of measurements at the RIMS level would be done so the reception at CPF level would be uniformly distributed. 

The analysis shows that for satellites with good observability (20 RIMS in view) 700 ms are required (with an acceptable probability) to detect “out of tolerance” events. Even with this optimistic hypothesis on satellite observability, the maximum TTA reduction obtained with asynchronous measurements is 300 ms. 

Increase RIMS Output Rate

In this concept, all RIMS will generate raw measurements at the same time but with a frequency of higher than 1Hz. First, the quality of these measurements must be verified and the detectable “out of tolerance” events analyzed. Figure 6 shows adopting a frequency up to 10Hz loop bandwidth for the PR measurements (compared to the usual 1Hz) would not have a significant impact on processing performance.

To take advantage of the increased RIMS output rate, the CPF integrity check algorithms shall be executed at the same frequency. To be consistent with the computation capability studied at CPF level, the proposed frequency is 4Hz. 

Taking into account the increased RIMS output rate of 4Hz and a CPF integrity check algorithm execution time of 250 ms, the SBAS message, including the alert, is ready to be broadcasted after 1,155 ms. Figure 7 shows the alert availability for broadcast.

The “waiting for next measurement” time is then reduced to 250 ms in the worst case, instead of the 1,000 ms for 1Hz measurement sampling. However, if the alerts are sent only through NOF with a 1Hz frequency, even if the system can detect alert conditions at a 4Hz frequency the alerts will be sent at 1Hz frequency. To take advantage of this improvement, the alerts shall be sent with the same frequency, 4Hz. 

Optimization of waiting for next broadcasting and broadcasting time

The alerts are sent in the NOF, which is broadcast with a 1Hz frequency synchronized with GPS time.

This generates a delay up to 1 second. Note the NOF is broadcast over a period of 1 second. Users can decode the alarm only after the complete reception. A delay of 1 second shall be added to the TTA.

This delay (up to 2 seconds) is directly linked to the alarm broadcasting solution.

To reduce this time, an alternative solution “fast alert message” has been analyzed to broadcast the alarm to users.

The aim is to send alerts to the user as soon as possible without waiting for the next SBAS message broadcasting, which is done one time per second synchronized with GPS time.

Fast alert messages contain the alert flags (alarm/no alarm) for all satellites set in the PRN mask. 

Management of fast alert messages implies a re-design of the user receiver that would require a MOPS standard update.

The user receiver shall consolidate the status of each satellite using both alert messages and NOF.

Figure 8 shows how the user receiver can incorporate fast alert messages to his NOF context.

“Construct NOF SIS” module constructs context using the NOF messages. After each alert message reception, the “extract alarms” module extracts the list of satellite alarms. The NOF context is then updated with the extracted alarms: a satellite is considered in alarm if an alarm is raised in fast alert message or in NOF SIS. The updated NOF context is then used like before.

Alarm hysteresis and repetition are ensured by the 1Hz chain in the NOF message.

There is no obligation to use the fast alert message for a receiver. There may be classes of users who can ignore the fast alert messages.

Fast Alert Message

SBAS alert messages, such as the Message Type 34, are transmitted today using a 250-bit long frame, including signalling and control fields (headers, CRC) at a transmission bit rate of 250 bits/s. As a consequence, a minimum 1-second latency is necessary to demodulate an alert message before processing the useful data. As an alternative, this article considers using the Q-component of the L1 and L5 frequency bands, with the strong assumption that the signal power available on the Q-component equals the I-component, without 
taking into account that other potential services could be broadcast on the Q-channel in the future. To reduce the alert message duration, and hence the TTA, two complementary approaches are considered: a bit rate increase requiring alternative signal waveforms and a shortened frame, at least for alert message types.

An analysis of the bitrate increase approach is provided in [2], where several potential candidate signal waveforms were proposed. In [2], the use of Code Shift Keying (CSK) modulation is particularly emphasized and allows for reaching an increased bit rate up to a factor 4, i.e., 1 kbit/s. The CSK was used in conjunction with specially designed Low-Density Parity-Check (LDPC) 
coding schemes within a Bit-Interleaved Coded Modulation, with Iterative Decoding (BICM-ID) architecture. The construction of GNSS waveforms bring additional arithmetical constraints on the sizes of LDPC uncoded words and coded words when taking into account the PRN sequence length, the number of code periods, and the size of the MT frame. In [2] the objective of the bit rate increase was reached, and the pros and cons of the candidate signals were proposed. However, using these alternative signals would constrain the alert message duration over 500 ms.

We preferred to use a more conservative approach, BPSK (because use of CSK would induce a heavy re-design of the receivers as it introduces a new modulation scheme) only as the Option 2B mentioned in [2], but with the perspective of a possible shorter alert message of 125 bits instead of 250 bits. It is foreseen that short-block LDPC codes have degraded performances with respect to larger block codes. Nevertheless, we envisaged to assess the Word Error Rate (WER) performance of an (125,250) LDPC code, concatenated with BPSK modulation, so as to enable 250 ms-latency alert messages. This proposal is presented in Table 2.

Some LDPC codes were proposed in an Experimental Specification published by CCSDS for short-block LDPC codes [3]. More specifically, we investigated the (128, 256) code and found a C/N₀
performance of ~30.7 dBHz for a WER=10-3, which is quite a bit over the 30 dBHz target.

An alternative to the CCSDS codes was proposed in [4], and the new parity-check matrix by Medova in the (128, 256) code has shown a 0.5 dB improvement, reaching a WER=10-3 performance for a C/N₀ of 30.25 dBHz, considered as an acceptable performance. Further improvements were also made to adapt the parity-check matrices to a (125, 250) code size without degradation of the C/N₀ performance. Even better LDPC codes can be designed to reach the required 30 dBHz, opening the path to 250 ms latency (alert) messages and the necessity to construct a 125-bit long message.

The new proposed message structure is based on MT34 from DFMC standard [5]. It is described in Table 3.

This message is used to transmit the Integrity Information through DFRECIs and DFREIs, for all the Augmented Slot Indices derived from the Satellite Mask. The payload is composed of 216 bits comprising: 92 DFRECIs (Change indicators), 7 DFREIs, Issue of Data Mask (IODM), 2 additional bits are reserved. The signalling and control (overhead) are composed of 34 bits: a preamble field, a Message Type ID field, and a Cyclic Redundancy Check (CRC) of 24 bits.

In the context of a shortened version of integrity information, we assume it shall have the lowest impact on other existing message types already transmitted on the I-component. This implies the satellite mask (MT 31), allowing to set up to 92 satellites, is maintained in its current definition. Hence, our main objective is to obtain a minimal alert information (1 bit for alert flag) for all 92 satellites within a total of 125 bits, which leaves only 33 bits.

After taking into account the IODM using 2 bits, the Message Type ID field over 6 bits, and the preamble over 4 bits, there remains only 21 bits for CRC. The standard CRC-24Q currently used in EGNOS is thus too long for this new potential framing. However, because the message frame length also has decreased, the choice for a shorter CRC was proposed. Our study shows it is possible to reduce the number of bits for the CRC with equivalent bit protection level (Hamming Distance), or without loss of integrity by replacing the CRC-24Q with a CRC-16 (0x9EB2).

Finally, we propose a new 125-bit message structure in Table 4 using a well performing CRC-16, as described in Table 3, compatible with the candidate signal Option 2D at either R_b=250 bits/s (500ms latency) or R_b=500 bits/s (250ms latency).

Fast Alert Message Broadcast

To shorten the “waiting for next broadcasting” time, the fast alert message can be sent as soon as an alert is available. However, if alert detection is based on 4Hz RIMS measurements, the fast alert message broadcasting can be done with a 4Hz frequency synchronized with alert availability at CPF output.

Fast alert message insertion makes it possible to eliminate the « waiting for next broadcasting » and to reduce the broadcasting time to 250 ms. The global delay is 500 ms instead of 2 s.

Conclusion 

This article identifies several options to improve the TTA for SBAS. 

It defines a first level of improvement that avoids any modification of the SBAS user standard and considers the benefit of optimizing CPF processing along with using measurements coming from wired RIMS only and/or additional RIMS measurements. These features allow us to reach a 4 second target compared with the current 6 second budget. 

A second level of improvement is obtained with a modification of the standard and the introduction of fast alert messages. Fast alert messages are broadcasted using the SBAS Q-channel and backward compatibility is maintained (Fast Alert messages could be ignored). This way, a target close to 2.5 seconds is met, as depicted in Figure 9. More details can be found in [6]. 

Acknowledgements

This work has been performed and funded under a contract of the European Space Agency in the frame of the EU Horizon 2020 Framework Program for Research and Innovation in Satellite Navigation.

Disclaimer 

The views presented represent the authors’ opinions. They should be considered R&D results and not taken to reflect the official opinion of the EU and/or the European Space Agency, in particular for what relates to present and future EGNOS system designs.

References 

(1) ICAO SARPs Annex 10 Volume 1 amendment 9

(2) Axel Javier Garcia Peña, Rémi Chauvat, Christophe Macabiau, Jaron Samson, Ivan Lapin, et al., Potential candidates for new SBAS signals. PLANS 2020 IEEE/ION Position, Location and Navigation Symposium, Apr 2020, Portland, United States

(3) CCSDS 231.1-O-1Short Block Length LDPC Codes for TC Synchronization and Channel Coding. Experimental Specification. Issue 1. Recommendation for Space Data System Standards (Orange Book)Book), CCSDS 231.1-O-1, Washington, D.C.: CCSDS, April 2015.

(4) L. R. Medova, P. S. Rybin and I. V. Filatov, “Short Length LDPC Code-Candidate for Satellite Control Channel,” 2018 Engineering and Telecommunication (EnT-MIPT), Moscow, Russia, 2018, pp. 163-166, doi: 10.1109/EnT-MIPT.2018.00044.

(5) ED259A (draft 21-06-2023) MOPS for Galileo/GPS/SBAS airborne equipment 

(6) C. Renazé, C. Bourga, M. Clergeaud Thales Alenia Space, Toulouse, France, J. Samson ESA. Reduction of system time to alert on SBAS. NAVITEC 2022 

Authors

Céline Renazé is a specialist in safety architecture at the Algorithms Design and Performance department of Navigation Domain, Thales Alenia Space. She received her M.S. in software engineering from the Paul Sabatier University, Toulouse, France, in 2003. She currently works on SBAS ASECNA as a System Architect.

Christophe Bourga is the head of the Navigation Systems Architecture department at Thales Alenia Space. Since 1997, he has worked on various topics in the navigation domain: GNSS mission, payload design, Galileo signal, formation flying, Galileo algorithms and performance, EGNOS architecture and evolutions, and ARAIM.

Matthieu Clergeaud is a system engineer at the Radionavigation department at Thales Alenia Space. He received his Engineering degree in Space Communications Systems from Telecom INT (Institut National des Télécommunications) in 2003. He now works on EGNOS-related projects.

Jaron Samson works as a Navigation Engineer at ESA’s Technical Centre (ESTEC) and has been involved in EGNOS, GNSS Evolutions, and R&D related to navigation. From 2012 until 2021 he worked as a System Engineer at ESA’s EGNOS Project Office in Toulouse, France. Jaron holds an M.Sc-degree in Physical and Mathematical Geodesy from the Delft University of Technology, The Netherlands.

The post Reduction of System Time to Alert on SBAS appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

A comparison against classic models.

SHISHIR PRIYADARSHI, WAHYUDIN P. SYAM, ANDRÉS ABELARDO GARCÍA ROQUÉ, ALEJANDRO PÉREZ CONESA, GMV

GUILLAUME BUSCARLET, RAÜL ORÚS PÉREZ, MICKAEL DALL’ ORSO, EUROPEAN SPACE AGENCY (ESA)

Sun is the source of energy needed to ionize the Earth’s atmosphere, which is an envelope of different gases. It has 11-year solar cycles where solar activity varies from low, medium to high [1]. Solar activity is generally measured using the sunspot number as well as solar radio flux (also known as F10.7 cm Radio Flux). Sunspot numbers are between 0 to 30 during low solar activity, 30 to 60 during medium solar activity and > 60 with the ability to reach up to 300 during high solar activity.

With increasing solar activity, the Sun emits stronger particles as well as radiation toward Earth through solar events such as Coronal Mass Ejections (CMEs) and solar flares. Solar activity is also recognizable from the Sun’s appearance as well as a terrestrial phenomenon known as Aurora. Solar radiation ionizes the upper atmospheric layer of the Earth between 80 to 600 km. Because of the dominant presence of ions and electrons, this layer is widely known as the ionosphere. The ionosphere overlaps the upper atmosphere to the region where space begins. In its top atmospheric layer, the gases are cooked by solar radiation until they emit a few electrons. This process creates a sea of ions and electrons within the ionosphere, allowing the process of ionization and recombination to keep running seamlessly. The ionosphere is a highly variable medium. The ionospheric irregularities that are a direct threat for the trans-ionospheric radio communication spatially varies from a few centimetres to 1,000 kilometres on special scale. 

Global infrastructure relies highly on a constellation of satellites (GNSS) that provides radio signals from space that transmit positioning and timing data to GNSS receivers [2-7]. These data are used for services like navigation, ranging, agriculture, tourism and transport. GNSS has global coverage, which is achieved using Europe’s Galileo, the U.S.’s Global Positioning System (GPS), Russia’s Global’naya Navigatsionnaya Sputnikovaya Sistema (GLONASS) and China’s BeiDou Navigation Satellite System. 

The ionosphere is highly turbulent on a temporal scale where significant changes are observed from fractions of seconds, hours, days, seasons, years and solar cycles. It’s challenging to model a turbulent medium like the ionosphere. There are well established state-of-the-art ionospheric models, but when compared with real data, they significantly deviate from the actual ionospheric condition checked using GNSS observations. Today, several daily life services such as transport, weather, banking transition, tourism and navigation are highly reliant on the Total Electron Content (TEC) fluctuations derived along the navigation path of the GNSS signals. [8-12]. Improving GNSS positioning is challenging, and neural networks are one-of-a-kind robust tools that are useful for that purpose [13]. One of the methods in trend for characterization of the ionosphere and navigation is GNSS high frequency signal analysis [14]. TEC is a descriptive measure of the electron quantity between GNSS satellite and receiver, which is estimated using the time delay between two GNSS signal frequencies. Users cover a wide range of the ionosphere using a signal dual frequency receiver [15-16]. TEC is often represented in the multiples of TEC unit, where 1 TECU = 1016 el/m2, which is about 1.66 X 10-8 mol.m-2. 

A few widely used and accepted empirical models such as Klobuchar [17], NeQuick-G [18] and International Reference Ionosphere (IRI) can produce the global morphology of the ionosphere during the geomagnetic quiet and disturbed days. Further, International GNSS Service (IGS) facilitates Global Ionospheric Maps (GIM) that records VTEC values of a spatial resolution of 2.5°×5°, and a 2-hour’s temporal resolution. These models and tools are good at reproducing the ionospheric geographic region as well as solar/geomagnetic activity dependent characterization, but still vary in error equivalent to 2-8 TECu for the same geographic locations during alike geomagnetic and solar activity condition. Geometrical effects are dominant at the lower latitude between satellite and receivers and add-on error to the ionospheric TEC estimation [19-20]. Physics based ionospheric models are often compromised in performance during low solar and geomagnetic activity [21-23]. These limitations and drawback of the ionospheric model and tools motivate the implementation of multi-layer machine learning (ML)-based models to improve the TEC forecasting accuracy and mitigate the degradation of the position accuracy induced by the ionosphere. Neural Network (NN) based TEC prediction models such as Han et al. [24], Ren et al. [25], Sivakrishna et al. [26], Shi et al. [27], Chen et al. [28], etc. are compromised as their predictive error increases with time and the global NN based TEC models fail to maintain regional accuracy in their forecast [29-30]. Another key aspect behind combining Multi-Layer Perceptron (MLP) + RNN, a hybrid approach, is the influence of other factors, such as space weather and geomagnetic indices, on the variations in the nonstationary patterns of ionospheric TEC.

This study compares the performance of the HARMONY (macHine leARning to MOdel gNss sYstems) ionospheric demonstrator (MLD-IONO) to the NeQuick-G, Klobuchar ionospheric models and GIM maps) during the high solar activity period at three different geographic locations. 

Data and Methodology

Machine Learning Ionospheric Model (MLD: ML IONO Demonstrator)

Using a combination of Multi-Layer Perceptron (MLP) and a Long-Short Term Memory (LSTM) based recurrent neural network (RNN) (Figure 1-1) a ML ionospheric model has been trained, which can nowcast and forecast the ionosphere over Europe and North Africa latitudes. While this abstraction seems intuitive and using the RNN can be considered because of its known capacity to process inputs of any length as well as to better “remember” the patterns along a time sequence, the applicability and benefits of this architecture is already determined during model validation. To assess the feasibility of this hybrid architecture, we considered a traditional RNN as well as a GRU and a LSTM. Model training used RINEX observation and navigation files along with the solar and geomagnetic indices, which were taken from public archives. The combination of MLP and LSTM keeps the good regression performance while providing additional capability to best exploit the information along the time axis. The final model has been tested and validated against real observations and along with state-of-the-art single frequency ionospheric models such as NeQuick-G and Klobuchar.

NeQuick-G is a three-dimensional time varying electron density model that predicts monthly mean electron density using the analytical profiles, geographic location and solar activity condition. Klobuchar is a simple structure model as it considers the ionosphere as a single layer at the height of 350 km. 

Ionospheric correction coefficients, which are eight in numbers, are transmitted by the satellites in the navigation message on a daily basis for GPS and are used to calculate the ionospheric delay. The Klobuchar model uses a 5 previous days average of solar flux due to which Klobuchar coefficients are updated once a day for GPS. These coefficients are based on the day of year (DOY). 

Figure 1-2 shows the location of stations considered for: (a) Ionosphere model implementation, (b) Training phase, (c) Validation phase, and (d) Training, testing and validation of the ionosphere model. Different experiments were performed using different configurations for the MLP architecture. Best hyperparameters were an MLP with 5 layers and an initial layer with 512 neurons and ReLU as the activation function, learning rate equal to 1e-5 and batch size of 184.

Because we decided to consider Solar cycle 24 as the period to collect our data and perform the analysis, we had to first understand how this solar cycle has evolved by looking at its periodicity. As illustrated in Figure 2, it is possible to distinguish an ascending and descending phase along the period from 2008 to 2020 characterized by high solar activity (high values of the F10.7 and SN parameters) during the ascending phase and lower solar activity during the descending phase.

It is worth noting that even in Figure 2, both geomagnetic and solar indices are represented, and the two groups of indices have a different scale and periodicity. With a closer look at their normalized evolution along the same time span (Figures 3 and 4), it is easier to distinguish how the solar parameters describe a clear ascending-descending cycle reaching peak during 2014, while geomagnetic parameters do not fluctuate following the same pattern.

Solar and geomagnetic parameters have been considered because of their influence on ionosphere behavior. When plotting the evolution of the STEC values (Figure 5), it is possible to see three different timescale patterns: 

1. SOLAR CYCLE VARIABILITY: This pattern matches one previously defined by the variation of the solar parameters. In this case, STEC values rise according to the solar cycle ascending phase to later decrease on the descending phase. Similarly, STEC peak values were obtained in 2014. Each STEC value has been computed as the mean of all the STEC values recorded for all the stations available for that particular point in time.

2. SEASONAL VARIABILITY: While you need to analyze the STEC along multiple years to visualize the first-time scale, it is also possible to find a second time pattern associated to the seasonal variability (Figure 6). Along each year, the STEC values show a steep increase during winter that progressively decrease during spring, summer and autumn. It is worth noting that by the end of spring and the start of summer, the STEC value shows a second smaller peak. As before, each STEC value is the averaged STEC of all available stations for that given point in time.

3. DAILY VARIABILITY: The third 
timescale, which influences STEC behavior, is associated with diurnal and nocturnal cycle. As seen in Figure 7, STEC values tend to reach peak during midday hours while remaining lower at night (early morning and late evening). STEC values represented in this plot are values of one fixed station and fixed date.

Results and Discussion 

Comparison of the MLD STEC/VTEC forecast to the known real STEC and GIS VTEC

High solar activity conditions are interesting to model at low-, mid-, and high-latitude regions due to the high solar indices that control the ionospheric instabilities formation dynamics. The presented space weather scenario covers the execution of the ML model for ionosphere modeling during a high ionospheric activity condition on March 8, 2014 for the low-, mid- and high- latitude stations Rabat Morocco (RABT00MAR), Ondrejov, Czechia (GOP600CZE), and Ny Alesund, Norway (NYA200NOR) respectively. 

The STEC and ∆STEC comparison at the three stations during high ionospheric activity that day (Figure 8) shows the MLD STEC output was much better than the NeQuick-G and Klobuchar modeled STEC at all three stations. Figures 8 and 9 demonstrate that MLD STEC and VTEC are close to the STEC observation and IONEX (IONosphere map Exchange format) VTEC whereas the MLD difference to the real data scenario seems to vary around zero, which means they have minimal deviation from the real data in comparison to the NeQuick-G and Klobuchar ionospheric model produced STEC/VTEC. IONEX format developed to exchange, compare or combine TEC maps. IONEX format supports the exchange of two- and three-dimensional TEC maps given on a geographic grid.

In this particular case, MLD VTEC performed better than NeQuick-G-G. The Klobuchar-produced VTEC matches quite close to the NeQuick-G VTEC (Figure 9).

Comparison of R2 STEC/VTEC 

Tables 1 and 2 show the MLD KPIs are better than the two models used at mid and high latitudes, however they are comparable to NeQuick-G at low latitude.

The correlation table shows the MLD performance was better overall (Tables 1 and 2) NeQuick-G’s performance was good for mid and high latitude, whereas Klobuchar VTEC has high value at mid- and low-latitude stations on March 8, 2014. 

MLD KPIs were best for STEC at all latitudes, for VTEC MLD is comparable to NeQuick-G at the station ‘GOPE.’ At the ‘RABT’ station, all three model outputs are comparable and within acceptable RMSE range on March 8, 2014. The NeQuick-G ionospheric model’s forecast of STEC and VTEC was executed for all three stations. We present here STEC data at a 20 to 40 degree elevation angle. The same input parameters have been used for shorting data for ML model predictions, as well as the NeQuick-G model. For VTEC, we provided ionospheric pierce point latitude and longitude at a shell height of 350 km.

Comparison of NeQuick-G/Klobuchar Ionosphere Model and STEC/VTEC predictions

The NeQuick-G-G/Klobuchar ionospheric model’s forecast of STEC and VTEC was executed for all three stations. We present STEC data at a 20 to 40 degree elevation angle. The same input parameters were used for shorting data for the ML model predictions, as well as the Klobuchar model. For VTEC, we provided ionospheric pierce point latitude and longitude at a shell height of 350 km (Figures 11 and 12).

GOP6 and NYA2 stations had the best MLD STEC and VTEC performances. But it was likely redundant to the NeQuick-G and Klobuchar for the ‘RABT’ station.

MLD KPI’s comparison to the NeQuick-G/Klobuchar ionospheric model 

The accuracy of the MLD-IONO model was tested using well established ionospheric models Klobuchar and NeQuick-G during low, medium and high solar activity. Overall, for high and mid latitude, MLD-IONO KPIs are lower than the OBS-NeQuick-G and OBS-Klobuchar KPIs. However, due to the lack of enough training data at the low latitude, MLD-IONO, performance is comparable to NeQuick-G and is better than the Klobuchar ionospheric model (Figures 11 and 12). 

Summary, conclusion and future works

The HARMONY MLD-IONO is intended to overcome the limitations of simulations based on a GNSS macro-model through the implementation of a ML model that models a GNSS system, providing better performances once trained to predict its behavior. Apart from a performance boost, the ML demonstrator could also leverage the use of previously unused data sources, providing knowledge of hidden phenomena and unknown relations. It was also found that the ML model predicted VTEC more realistically than NeQuick-G. Klobuchar forecasted VTEC matches quite close to NeQuick-G. The performance of the developed ionospheric machine learning model is highly dependent on solar and geomagnetic activity and the receiver location.

In general, with all the stations, demonstrator performance was within the acceptable KPIs range. In the future, we would like to extend this study by ingesting ground magnetometer data to see any possible connection between a ground induced current and the ionospheric variability. Our future activities will focus on developing new compact tools and Sun-Earth relationships based on space weather models suitable to mimic/forecast the extreme solar events that are a threat to the power grid and services that rely on it. 

Acknowledgments

This work was supported by the European Space Agency NAVISP program and funded under the activity EL1-035 ter: “Machine Learning to Model GNSS Systems.”

References

(1) Nicola, S., Antonio, B. (2022). The Planetary Theory of Solar Activity Variability: A Review, Frontiers in Astronomy and Space Sciences, 9(2022), https://doi.org/10.3389/fspas.2022.937930

(2) Morton, Y., Taylor, S., Wang, J., Jiao, Y., Pelgrum, W. (2013). In L band ionosphere scintillation impact on GNSS receivers. In Proceedings of the IEEE Radio Science Meeting (USNC-URSI NRSM), US National Committee of URSI National, Boulder, CT, USA, 9–12 January 2013, 1.

(3) Chen, X., Dovis, F., Peng, S., & Morton, Y. (2013). Comparative studies of GPS multipath mitigation methods performance. IEEE Transactions on Aerospace and Electronic Systems, 49(3), 1555–1568.

(4) Ghafoori, F., Skone, S., (2015). Impact of equatorial ionospheric irregularities on GNSS receivers using real and synthetic scintillation signals. Rad. Sci., 50, 294–317.

(5) Kintner, P., Ledvina, B., De Paula, E. (2007). GPS and ionospheric scintillations. Space Weather 5, 83–104.

(6) Jacobsen, K.S., Dahnn, M. (2014). Statistics of ionospheric disturbances and their correlation with GNSS positioning errors at high latitudes. J. Space Weather Space Clim., 4, A27.

(7) Tiwari, R., Strangeways, H.J., Tiwari, S., Ahemad, A. (2013), Investigation of ionospheric irregularities and scintillation using TEC in high latitude. Adv. Space Res., 52, 1111–1124.

(8) Tang, L., Yang, X., Kan, Z., Li, Q. (2015). Lane-level road information mining from vehicle GPS trajectories based on naïve bayesian classification. ISPRS Int. J. Geo-Inf., 4, 2660–2680.

(9) Xie, X., Wong, K.B.-Y., Aghajan, H., Veelaert, P., Philips, W. (2015). Inferring directed road networks from GPS traces by track alignment. ISPRS Int. J. Geo-Inf., 4, 2446–2471.

(10) Li, J., Zhang, Y., Wang, X., Qin, Q., Wei, Z., Li, J. (2016). Application of GPS trajectory data for investigating the interaction between human activity and landscape pattern: A case study of the Lijiang River Basin, China. ISPRS Int. J. Geo-Inf., 5, 104.

(11) Ranacher, P., Brunauer, R., van der Spek, S., Reich, S. (2016). What is an appropriate temporal sampling rate to record floating car data with a GPS? ISPRS Int. J. Geo-Inf., 5, 1.

(12) Wu, T., Xiang, L., Gong, J. (2016). Updating Road networks by local renewal from GPS trajectories. ISPRS Int. J. Geo-Inf., 5, 163. 

(13) Kanhere, A. V., Gupta, S., Shetty, A., and Gao, G. (2022). Improving GNSS Positioning Using Neural-Network-Based Corrections. NAVIGATION: Journal of the Institute of Navigation, 69 (4) navi.548. https://doi.org/10.33012/navi.548

(14) Baumgarten, Y., Psiaki, M.L.,and Hysell, D.L. (2022). Navigation and Ionosphere Characterization Using High-Frequency Signals: A Performance Analysis. NAVIGATION: Journal of the Institute of Navigation, 69(4) navi.546. https://doi.org/10.33012/navi.546

(15) Farzaneh, S., Forootan, E. (2020). A Least Squares Solution to Regionalize VTEC Estimates for Positioning Applications. Remote Sens., 12, 3545. https://doi.org/10.3390/rs12213545

(16) Ya’acob, N., Abdullah, M., Ismail, M. (2010). GPS Total Electron Content (TEC) Prediction at Ionosphere Layer over the Equatorial Region. In: Bouras, C., editor. Trends in Telecommunications Technologies [Internet]. London: IntechOpen; 2010 [cited 2022 Nov 25]. Available from: https://www.intechopen.com/chapters/9691. https://doi.org/10.5772/8474

(17) Klobuchar, J. (1987). Ionospheric Time-Delay Algorithms for Single-Frequency GPS Users. IEEE Transactions on Aerospace and Electronic Systems, 3, 325-331.

(18) Prieto-Cerdeira, R., Orús-Pérez, R., Breeuwer, E., Lucas-Rodríguez, R., and Falcone, M. (2014). ‘Innovation: The European Way. Performance of the Galileo Single-Frequency Ionospheric Correction During In-Orbit Validation’. In: GPSworld 25.6, 53–58. Categories: FundamentalsGNSS Measurements Modelling. 

(19) Singleton, D.G. (1970). Dependence of satellite scintillations on zenith angle and azimuth. J. Atmos. Terr. Phys. 32, 5, 789-803, https://doi.org/10.1016/0021-9169 (70)90029-2

(20) Priyadarshi, S., Wernik, A.W. (2013). Variation of the ionospheric scintillation index with elevation angle of the transmitter. Acta Geophys. 61, 1279–1288. https://doi.org/10.2478/s11600-013-0123-3

(21) Priyadarshi, S. (2015a). A Review of Ionospheric Scintillation Models. Surv Geophys, 36, 295–324. https://doi.org/10.1007/s10712-015-9319-1

(22) Priyadarshi, S. (2015b). Ionospheric scintillation modeling for high- and mid-latitude using B-spline technique. Astrophys Space Sci., 359, 12. https://doi.org/10.1007/s10509-015-2461-x

(23) Priyadarshi, S., Zhang, Q.-H., Ma, Y.-Z., Wang, Y., and Xing, Z.-Y. (2016). Observations and modeling of ionospheric scintillations at South Pole during six X-class solar flares in 2013. J. Geophys. Res. Space Physics, 121, 5737– 5751. https://doi.org/10.1002/2016JA022833

(24) Han, Y., Wang, L., Fu, W., Zhou, H., Li, T., and Chen, R. (2021). Machine Learning-Based Short-Term GPS TEC Forecasting During High Solar Activity and Magnetic Storm Periods. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 15,115-126. https://doi.org/10.1109/JSTARS.2021.3132049

(25) Ren, X., Yang, P., Liu, H., Chen, J., Liu, W. (2022). Deep Learning for Global Ionospheric TEC Forecasting: Different Approaches and Validation. Space Weather, 20, e2021SW003011.

(26) Sivakrishna, K., Venkata, R.D., Sivavaraprasad, G. (2022). A Bidirectional Deep-Learning Algorithm to Forecast Regional Ionospheric TEC Maps. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., 15, 4531–4543.

(27) Shi, S., Zhang, K.,Wu, S., Shi, J., Hu, A., Wu, H., Li, Y. (2022). An Investigation of Ionospheric TEC Prediction Maps Over China Using Bidirectional Long Short-Term Memory Method. Space Weather, 20, e2022SW003103.

(28) Chen, Z., Liao, W., Li, H., Wang, J., Deng, X., Hong, S. (2022) Prediction of Global Ionospheric TEC Based on Deep Learning. Space Weather, 20, e2021SW002854.

(29) Priyadarshi, S., Syam, W., Roqué, A.A.G., Conesa, A.P., Buscarlet, G., Pérez, R.O. and Orso, M.D. (2023, January). Machine Learning-based ionospheric modelling performance during high ionospheric activity. In Proceedings of the 2023 International Technical Meeting of The Institute of Navigation (ION ITM 2023) (pp. 950-965).

(30) Syam, W.P., Scott, D., Perez-Conesa, A., Rodriguez, I., Lechuga, M.L., Martinez, E.J. and Ioannidis, R.T. (2023). Navigating Emergencies with a Low-RF CARS. Inside GNSS, 18, 1, p.40.

Authors

Shishir Priyadarshi obtained his Ph.D. in Space Physics from the Space Research Centre, Poland, in 2014. Thereafter, he held various international postdoctoral, visiting scientist and adjunct faculty positions between 2014 and 2021. Subsequently, he joined GMV as a Machine Learning Engineer in 2022.

Wahyudin P. Syam holds a Ph.D. in geometrical measurement and uncertainty analysis from Politecnico di Milano in Milan, Italy. He joined GMV in February 2021. At GMV, he is involved in several ESA projects related to GNSS fingerprinting and IGS orbit and clock-bias corrections.

Andrés García holds a bachelor’s degree in computer engineering as well as a master’s degree in data science and engineering. For two years he has worked in GMV Innovating Solutions as a big data engineer and then changed to a machine learning engineer position.

Alejandro Pérez Conesa received his B.Sc. degree in Computer Engineering and his B.Sc. and M.Sc. degrees in Telecommunication Engineering from the Universitat Autònoma de Barcelona (UAB) in 2018, 2019 and 2020, respectively. He is currently working as Project Manager for GMV and as Adjunct Professor at UAB. 

Guillaume Buscarlet graduated from Supaero (ENSAE) Toulouse, France, in 1997, worked at Thales Alenia Space, Toulouse, France in Navigation (EGNOS) and Telecom (IRT Saint-Exupery, Digital P/L). Since 2020, he has worked at ESA (EGNOS Project Office) as a SBAS System Performance Engineer for EGNOS.

Raul Orus Perez received a Ph.D. in Aerospace Science and Technology from the gAGE/UPC research group of the Technical University of Catalonia (UPC) in 2005. Since 2010, he has worked at ESA/ESTEC as a Propagation Engineer (radio-wave propagation in troposphere and ionosphere).

Mickael Dall’Orso received his M.Sc degree from ISAE-ENSICA (French Institute of Aeronautics and Space) in 2010, Toulouse, France. He worked until 2018 in SBAS algorithms and performances (EGNOS, KASS) at Thales Alenia Space in Toulouse. He is currently an EGNOS System Performance Engineer at ESA’s EGNOS Project Office in Toulouse.

The post High-Solar Activity Ionospheric Modeling Leveraging Machine Learning appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

Guest contributors Darioosh Naderi and Mark Marshall of Silicon Sensing look at how far the technology has come and why it’s poised to disrupt the market.

For years, manufacturers have focused on the consumer and industrial grade performance that MEMS inertial technology provides, with many in the industry assuming that’s as far as MEMS capabilities can take them. That perception is beginning to shift, though, with several companies proving their MEMS inertial sensors are capable of achieving tactical performance levels—and that’s changing the game.

When users switch to high-performance MEMS sensors, they’re often surprised with the results, Silicon Sensing’s Head of Business Development Darioosh Naderi said. The level of performance provided goes well beyond their expectations, with the solutions displacing fiber optics and other technologies, while demonstrating they can perform exceptionally well over a wide range of environments. 

As strong as it is today, MEMS certainly hasn’t reached its limits. The technology has so much potential in terms of the capabilities and performance it can provide in the future, helping to solve many of the challenges the industry faces. Companies like Silicon Sensing are working to advance MEMS technology and unlock its full potential.

“There’s a state of flux right now, and companies are starting to go further with the technology,” Silicon Sensing Chief Engineer Mark Marshall said. “The accepted benefits of MEMS really are their lower size, weight and power consumption—as well as their potential for mass manufacturing, which keeps the cost lower than traditional inertial technology. Alongside these, two lesser-known benefits are their robustness and reliability—a consequence of the lack of moving parts in each unit. This means MEMS products will maintain performance in very harsh environments.” 

Marshall continues: “There is no doubt there remains huge, untapped potential in MEMS-based technology, which has been limited to date by its ability to deliver market-disruptive performance.”

That disruption, however, is on the horizon as technological advances mean a number of inertial sensors are now demonstrating navigation grade performance. 

How MEMS Compares

There are various technologies that can be leveraged in inertial systems, all with pros and cons. “Mechanical gyros, for example, have been around for years and are a well understood, mature technology,” Marshall said. “But because they have moving parts, they need more maintenance, cost more, and aren’t as robust as MEMS alternatives.”

“Mechanical systems typically need to be recalibrated and must go through a certification period,” Naderi explained. “They may also require extensive maintenance, such as work to replace complex, well-balanced bearings.”

“Ring laser gyros (RLG) have advantages over mechanical systems,” Marshall continued, “with the biggest being their precision performance. RLG unit size can be an issue, but they are becoming smaller as manufacturing processes improve. However, they are not the most affordable solution. The cost of these devices is much higher than both mechanical systems and MEMS.”

“Fiber optic gyros (FOG), another option, have an inherently high bandwidth that addresses noise issues and delivers excellent performance. But, again, size can be an issue because it is necessary to wind a longer coil to achieve higher performance levels. They also have a higher price point.”

As the performance of MEMS inertial sensors continues to advance, this technology is now starting to offer the capability and performance levels needed by applications with much stricter stability and precision requirements. This is beginning to render them a viable and affordable inertial option, even for applications that call for tactical grade performance, and beyond. 

Getting to Navigation Grade 

“Over the last 20 years, MEMS inertial sensors have continued to improve,” Marshall explained, “with improvements made in a variety of areas including bias, drift, noise, scale factors and angle random walk.” MEMS inertial sensors that have traditionally been used for industrial grade applications are finding their way into more tactical and navigation-grade applications that demand higher precision and lower drift rates.

“There are incremental improvements available today,” Marshall continued. “But we’re also focusing on that stepwise improvement, that order of magnitude improvement, that will see MEMS inertial sensors truly disrupt the market for higher-end technologies. Silicon Sensing is just one of many companies with a focus on working toward navigation grade inertial sensors at reduced cost, size weight and power (CSWaP) with encouraging levels of success.” 

“For companies like Silicon Sensing, the real goal is to enable our partners to solve their guidance or navigation challenges, whilst recognizing the need to work with increased levels of signal disruption, an increased need for prolonged autonomy, and a requirement to operate in more extreme environments across land, air, sea and space.”

It basically comes down to solving two problems: Enabling users to gyrocompass in an often intense, dynamic environment, and providing aided navigation in a frequently GNSS denied environment.

“Today, you can take a Silicon Sensing MEMS sensor and rotate it around a step motor to find true north,” Naderi explained. “This is ideal for users, such as the downhole drilling market, who need to know where true north is quickly and unaided.” 

“Right now, companies are paying tens of thousands of dollars for complex and costly systems that allow them to gyrocompass when in reality they might not need all the performance their chosen solution provides,” he said. “Similarly, they may not be able to use a magnetometer because, while the performance may be sufficient, it may be too susceptible to local magnetic fields. The ability to provide a device that is capable of true north seeking without external aiding, doesn’t suffer from local interference, and has the size and price of a MEMS unit will significantly disrupt the market.” 

“To solve the gyro-compassing problem, solutions need to offer sub-one degree an hour for residual bias,” Naderi explained. “But for navigation-grade performance, you have to get much lower, again with different markets and applications having different requirements in terms of performance.” 

“The main goal is to get bias and associated parameters as low as possible. This will give us that order of magnitude ‘stepwise’ improvement Mark talks about needing to achieve,” Naderi explained. “Improving noise and scale factor are also priorities.” 

“And that’s where you say: What problems are you trying to solve and what are your key parameters. Most will say they are scale factor, bias and angle random walk,” Marshall said. “If you look at our gyro technology as standalone sensors with their own electronics, we are closing the gap to navigation grade in many parameters today. Some of our customers are virtually there without any aiding sources and without any mechanical rotation or gyrocompass. They are about to cross that threshold. But that’s just the gyros. There are also accelerometers to consider.”

Extremes of temperature is another key consideration. 

“It’s not just sitting on the bridge of a vessel at 25 degrees Celsius forever,” Marshall said. “Getting the sensors to work in any environment, from minus 55 degrees all the way up to plus 85 degrees, is a real challenge. We have customers operating at all temperature brackets—and in even more extreme situations.” 

A Look at the Market

Today, the main customers relying on tactical MEMS for an IMU are the surveying and mapping markets, 
marine gyro-compassing customers, and defense applications for guided munitions, Naderi explained. “One of the biggest markets primed for disruption is avionics. The performance parameters that must be met are clear, and while this won’t be easy, it is within reach for MEMS manufacturers.” 

The aerospace market is an obvious candidate for MEMS, Naderi believes, along with broader defense applications such as tactical wheeled vehicles and any theater of conflict where GNSS jamming and spoofing are concerns. In fact, MEMS would have application almost anywhere a ring laser gyro is currently in use. 

And because MEMS technology is a fraction of the price and size of other options, “it opens doors to applications tomorrow that we don’t even know exist today,” Naderi believes. 

“We could be creating new applications. Just one area with huge potential would be in the growing unmanned markets,” Naderi said. “Longer endurance and beyond visual line of sight is becoming more feasible. And when you have a navigation system that doesn’t rely on a constant external input, it presents so many new opportunities. It disrupts by definition.” 

Advancing the Technology

Silicon Sensing already has a patented high-performance gyro that has gone through multiple design iterations. Just one area the company is focused on today is material use, Marshall explained. “A key avenue we’re looking at is new novel materials in the MEMS structure itself. That’s something we believe there’s quite a bit of mileage in. There’s really some exciting work going on.” 

This includes looking at the detailed construction of the gyro and how the layers are made up, exploring re-positioning the MEMS within another structure with the ASICs to control them, and working on complex discreet electronics to drive the gyros optimally such as improving how information outputs are delivered. The company is also making advances in the way sensors are calibrated. 

“So, we are looking at a whole system on every front and we’re seeing many innovations we can introduce,” Marshall said. 

Critically, the company is also exploring improvements for more efficient volume production. 

Silicon Sensing has three different modes of operation for its gyros, Naderi explained. Some are driven with magnets, some are capacitive and others use a thin film PZT material. 

“If we look at the inductive gyro, we can improve the magnetic circuitry in it and gain benefits in the scale factor almost immediately,” Naderi said. “If we look at the design of the gyro, which is on a base, and if we change the base without changing the fundamental design, that gives us benefits in terms of bias. So, we can take the exact existing design, existing manufacturing process, make a minor tweak, and that can give, in some cases, a three-fold improvement. And we can do that today.”

“MEMS feature layers of material, including silicon, metal and glass,” Naderi said. “The thermal characteristics of these different layers is a key area of consideration. Manufacturers can look at different techniques to try to improve on existing designs without going back to the drawing board.” 

“You look at the improvements we can make to an existing drive type, with the magnet and with the structure it’s on, then you look at the material mismatching, so already you’ve got some ground,” Naderi said. “Then you look at the capacitive type gyros that are three millimeters or four millimeters, we can make them larger, and possibly apply some very clever real-time balancing techniques. And we can get a sense of a performance improvement that could be four times better than we have today. Versions of these sensors have been built and tested. Our focus now is on bringing that product to market with a set of electronics that will get the best out of the sensor, whilst maintaining the attractive size and cost of a typical silicon MEMS inertial product.”

Steering the Development

Silicon Sensing has its own MEMS foundry, enabling the company to produce parts at a much lower cost and at higher volumes. It also allows them to be far more involved in the design and manufacturing than those who outsource production. 

“We have the flexibility to be able to tweak the designs and we work very closely with the foundry processing, calibration and engineering teams. We have ultimate control,” Marshall explained, “so we can do all the finite element modeling and we can steer the development.” 

The foundry’s sole purpose is to turn out MEMS inertial sensors, Naderi explained, so the team has acquired a wealth of knowledge about the products that a third-party foundry just wouldn’t have. 

“It could take years to get a prototype that ticks all the boxes out of a third-party foundry,” Naderi said. “Our foundry has a very quick turnaround, speeding up product development and time to market. And it is a powerful asset, having decades of knowledge and foundry expertise.”

Looking Ahead

In the next three to five years, Naderi expects to see more companies pushing toward navigation grade MEMS offerings, with those small, iterative steps making revolution possible. The disruption these sensors will cause is closer to five years out. It comes down to putting everything together in a package that delivers what’s required and costs significantly less than alternative technologies. Accelerometers are also advancing. Silicon Sensing is working on offerings that have the required performance, with some promising initial results. However, formal qualification programs take time, as do the vital updates to production lines that will allow reliable production of these much higher performing sensors. 

“In some respects, the design of a navigation grade MEMS gyro exists today. Bringing it all together in terms of the drive electronics and reliable, full-scale production is our main challenge,” Naderi concluded. 

Authors

Dr. Mark Marshall was appointed Chief Engineer of Silicon Sensing Systems in February 2023 having joined Silicon Sensing in 2017 as the lead engineer on numerous inertial programs. Previously, Dr. Marshall worked as a senior research engineer in the field of laparoscopic electro-surgical instruments, with many patents and refereed papers. He has a first class honors degree in Computer Aided Engineering, an MSc in Advanced Manufacturing Systems and a PhD from the University of Cambridge.

Darioosh Naderi is the sales, marketing and business development lead for Silicon Sensing Systems Ltd (SSS). SSS is a leading provider of silicon micro electro-mechanical systems (MEMS)-based inertial sensors and systems, including gyros, inertial measurement units (IMU) and accelerometers. Mr. Naderi has worked in business development and product strategy within Silicon Sensing, and within one of its parent companies, for some five years.

The post MEMS: Unlocking the Potential appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

Identifying multipath reflectors for validating ray tracing in urban canyons.

PENG XIE, OPENLOOPNAV INC.

Global Navigation Satellite Systems (GNSS) have been widely used in many applications. Multipath mitigation in urban canyons is one of the last unsolved problems in GNSS as discussed in [1]. A large number of multipath signals can be observed in dense urban canyons such as downtown San Francisco. The receiver can be surrounded by many tall buildings, causing signals to bounce around among different buildings. Multipath causes the presence of several correlation peaks, which are computed by the receiver for estimating the signal parameters.

Clear examples are shown in Figure 1, where at least three correlation peaks are observed for a GPS satellite (left) and at least four correlation peaks for a Galileo satellite (right) with different delays and Doppler frequencies. Each peak corresponds to a signal component, either line-of-sight (LOS) or multipath. The red curves are the projections of the correlation peaks in the code and Doppler domain, respectively.

Shadow matching and 3D ray tracing techniques were proposed recently to improve position performance in urban canyons. Shadow matching uses a 3D city model to find the most likely user location by comparing received signal power and predicted satellite visibility [2-4 ]. Ray tracing is a technique that uses 3D city and digital elevation models to determine reflection points of GNSS signals between a given satellite and receiver position [1, 5].

The relationship between multipath occurrence and reflector size is analyzed by using the Fresnel zone concept. By determining possible reflection points on the reflecting surfaces, reflected signal amplitudes and ranges can be identified from a geometrical perspective [6]. The ray tracing technique works based on 
several assumptions, e.g., shape of reflector, smoothness of the reflector, materials, specular reflection or diffraction, etc. Usually, only a single reflection is assumed. In [5], Strandjord experimentally showed that, using current smartphones, it is not possible to establish whether the signals tracked are LOS or non-line of sight (NLOS) components. Note that in this context we classify both NLOS reception and multipath reception as multipath, even though the receiver may be affected differently by NLOS and multipath [3,4].

Determining how to evaluate ray tracing performance and validate assumptions used in ray tracing is still a challenge. Conventional methods reported in literature to evaluate ray tracing performance are based on signal power, range and Doppler errors [7]. In [8], Zhang et al. used single/double difference to evaluate ray tracing performance. Each method has its own limitations and often external aiding information is required. In this work, a standalone GNSS solution is proposed to better isolate signals observed in urban canyons. By isolating the LOS from multipath signals, accurate position solutions can be obtained in challenging environments such as downtown San Francisco. While the technique can be used for accurate navigation, it also provides an effective tool to assess the quality of ray tracing results.

A software receiver was designed for this purpose, and a digital front-end has been used to collect Radio Frequency (RF) data. In our architecture, a correlation map similar to the one shown in Figure 1 is available for every satellite at each epoch, meaning we have the ability to detect several rays at the same time. We use advanced signal processing and position estimation techniques to characterize each ray tracked. By using L1 signals (GPS L1, Galileo E1 and Beidou B1C), everything is resolved in the position engine, which features a Kalman filter architecture. Range errors of each ray are directly available and can be used to detect the reflectors causing multipath.

Doppler Frequency

Doppler frequency is the signal frequency change from satellite to receiver due to relative motion between them. The Doppler frequency perceived by the receiver is comprised of contributions from satellite motion, receiver motion and receiver clock effects. LOS signal Doppler can be written as

where  is the satellite velocity vector,  is the receiver velocity vector,  is the unit LOS vector, d_R is the receiver clock drift, and λ is the signal wavelength. Similarly, multipath signal Doppler can be written as

where  is the multipath unit vector.

Previous research showed that for dynamic scenarios, LOS and multipath signal generally have different Doppler frequencies [9, 10]. Multipath Doppler difference can be written as

In a 2D scenario, if we know the multipath Doppler error, df, receiver velocity, , and LOS vector, , we can obtain the angle θ between  and , which is the multipath direction:

However, θ is a combination of elevation and azimuth in a 3D coordinate system, so the multipath vector  could be anywhere on a cone surface as shown in Figure 2. If we assume multipath elevation is the same as LOS elevation, the multipath azimuth ambiguity can be reduced to two values, i.e., left side and right side on the cone surface with the same elevation. 

Ray Tracing and Fresnel Zone

Ray tracing is commonly used in computer graphics for image synthesis. The light propagation path from its source is traced as it bounces multiple times around the scene. In this work, ray tracing refers to an approach that uses a 3D city model to analyze the signal path between a given satellite and the receiver position. By determining the reflection points on the obstruction surfaces, possible reflected paths can be identified. In the case of reflection, there are many areas on the obstructing surface that contribute to the reflected signal. Those regions are named Fresnel zones [6]. The Fresnel ellipsoid of the first-order has the most energy transmitted; in this regard, the first-order Fresnel zone is used to decide if reflection is significant enough to affect a receiver. 

As shown in Figure 3, the Fresnel zone is overlapped with the 3D building model, where satellite position is a focal point of the ellipsoid. In the figure, the small yellow circle on the left side of the blue building represents the receiver, whereas the yellow circle on the right side represents the mirror of the receiver onto the reflector plane (another focal point of the ellipsoid). R is the reflection point. Usually, if the overlapping is more than 50% of the Fresnel zone, it is assumed that the reflected signal will affect receiver performance [5,6].

A similar idea can be applied to LOS propagation. LOS Fresnel zone is defined in Figure 4 and closely following the detailed description documented in [11], the radius R_f of the Fresnel zone at any point along the path is given by:

where D_s is the distance between satellite and point D, and D_R is the distance between receiver and point D. LOS signal power is decided by this Fresnel zone. If there is no obstruction between 
satellite and receiver, we will get a nice correlation peak. However, if some obstructions are present on the propagation path and overlap with the Fresnel zone, the LOS signal will be diffracted at the edges of the object; consequently, the LOS power will be attenuated.

The power of LOS and multipath can change quickly during a short period of time in urban canyons, especially for driving scenarios. This depends on the overlap between the Fresnel zone and obstructions.

An illustration of this phenomenon is shown in Figure 5 downtown San Francisco. A LOS peak and a strong multipath peak can be observed for GPS Satellite Vehicle (SV) 8 with 41 degrees elevation, 118 degrees azimuth. Correlation plots are shown for three consecutive seconds in Figure 5: the building highlighted in the red rectangle can partially block the LOS signal. On the top left plot of the figure, the LOS peak (in the middle) is weaker compared to the multipath (peak on the far left side); after one second (top right plot), the LOS peak (in the middle) is much stronger compared to one second earlier.

The LOS power variation during these consecutive two seconds is due to the Fresnel zone discussed earlier, which is also shown in the bottom right plot, where the overlapping between LOS Fresnel zone and the obstruction is changing rapidly. The yellow circle corresponds to the top left plot, with more than 70% of the Fresnel zone overlapped with the building, which is why the LOS signal is very weak; the blue circle corresponds to the top right plot, with less than 50% of Fresnel zone overlapped with the building, thus, the LOS signal is stronger compared to the previous second. Finally, the green circle corresponds to the bottom left plot, where there is no overlap between the Fresnel zone and the building. Thus, a strong LOS peak can be observed.

A commercial receiver may track the overlapped version of correlation peaks, depending on how the receiver is designed.

Reflector Direction Prediction

Using the approach developed we can tell from which direction the multipath component is originating. This section discusses how to calculate the reflector direction based on several assumptions when the receiver is moving. This can be used to evaluate ray tracing performance in several ways to answer the following questions:

• Is multipath direction and signal power observed the same as the ray tracing prediction?

• Is range error and/or Doppler error observed the same as the ray tracing prediction?

• Is the ray number observed the same as the ray tracing prediction?

To demonstrate these points, we collected data from the city of San Jose. Red dots in Figure 6 show the field test route. We selected two locations (A and B) to study how to predict reflector direction from our solution.

GPS SV 16 is used in this study, elevation is 27.7 degrees, and azimuth is 288.7 degrees. Receiver position at location A is [37.332472 -121.893792], and speed is 6.6 m/s. Figure 7 shows GPS SV 16 L1 signal correlation peaks at location A.

Table 1 summarizes the correlation peaks shown in Figure 7.

From Equation 4, we can get θ is 48.5 degrees for ray 2, which is a multipath with 8.9 m/s Doppler error (i.e., df=8.9). If we assume it is a specular reflection and multipath elevation is also 27.7 degrees (the same as LOS), we can obtain two azimuths of multipath direction (106 degrees and 189 degrees), these two directions are on the same cone surface as depicted in Figure 2. Now we can do a simple ray tracing based on Google Earth to evaluate our prediction. From Google Earth two possible reflection points, R1 (azimuth 189 degrees) and R2 (azimuth 106 degrees), can be identified as shown in Figure 8. Green arrows are LOS vectors, yellow arrows are multipath vectors and the red arrow is velocity vector. Reflector information is summarized in Table 2.

After calculating the geometry range from point R1 and R2 to location A separately, we conclude that point R1 should be the correct reflector. The extra path from reflector R1 is 91 m, which matches the range error shown in Table 1 (91 m versus 90.5 m). This confirms our reflector direction prediction is reliable, i.e., we can actually find a valid reflector from this direction.

After 400 s, a similar multipath peak showed up at location B as depicted in Figure 9. Receiver position at location B is [37.329719 -121.891592], and receiver speed is 8.0 m/s. Table 3 summarizes correlation peaks observed from this figure. Again, two possible multipath azimuths can be obtained in Table 4, i.e., 189 degrees and 107 degrees. From Google Earth only one possible reflection point R1 (azimuth 189 degrees) can be identified as shown in Figure 10; no valid reflection point can be observed from 107 degrees.

After calculating the geometry range from point R1 to location B, we conclude that point R1 should be the correct reflector. The extra path from reflector R1 is 94 m, which matches the range error shown in Table 3 (94 m vs. 87.7m).

From the case studies, we learned the solution we proposed in this work to evaluate ray tracing performance is accurate and reliable. Moreover, if the L5 signal is used, the range accuracy can be further improved as the L5 signal has higher range resolution.

Conclusion

In summary, we proposed a solution that can be used to evaluate ray tracing performance in urban canyons, which takes advantage of multipath Doppler behavior under dynamics. The solution we proposed is a standalone system without any aiding information used; moreover, it works in a reverse way to ray tracing, i.e., we don’t do ray tracing, but we can tell which ray is observed in the receiver, and the possible direction of each ray. The Fresnel zone concept is illustrated in this work, and can be used to evaluate assumptions made in ray tracing, e.g., building model, material permittivity, etc., which will affect the received signal amplitude in the receiver side. Two case studies in San Jose are conducted to verify our method, with results showing our solution is accurate and reliable to evaluate ray tracing performance. 

Manufacturers

The digital front-end used in the experimental analysis is a Universal Software Radio Peripheral (USRP) B200mini from Ettus Research.

References

(1) Van Diggelen, F., “End Game for Urban GNSS: Google’s Use of 3D Building Models”. Inside GNSS Magazine, March 2021

(2) Wang, L., Groves, P., and Ziebart, M., “GNSS shadow matching: Improving urban positioning accuracy using a 3D city model with optimized visibility prediction scoring”. In Proceedings of the 25th ITM of the Satellite Division of The Institute of Navigation (ION GNSS 2012), pp. 423–437

(3) Groves, P., “Multipath vs. NLOS Signals”. Inside GNSS magazine, November 2013

(4) Groves, P., “Shadow matching: A new GNSS positioning technique for urban canyons”. The Journal of Navigation. 2011, 64(3), 417– 430. https://doi.org/10.1017/S0373463311000087

(5) Strandjord, K. L., Axelrad, P., Mohiuddin, S., “Improved urban navigation with shadow matching and specular matching”. NAVIGATION. 2020;67: 547–565. https://doi.org/10.1002/navi.378

(6) Zimmermann, F., Schmitz, B., Klingbeil, L., and Kuhlmann, H., “GPS multipath analysis using fresnel zones”. Sensors. 2018, 19(1), 25. https://doi.org/10.3390/s19010025

(7) Panicciari, T., Soliman, M. A., Moura, G., “Simulating multipath in real time for receiver evaluation”, GPS World, March 2018

(8) Zhang, G., Xu, B., Ng, H.F., Hsu, L.T.,“GNSS RUMS: GNSS Realistic Urban Multiagent Simulator for Collaborative Positioning Research”. Remote Sens. 2021, 13, 544. https://doi.org/10.3390/rs13040544

(9) O´Driscoll, C., Lachapelle, G., and Tamazin, M., “Dynamic Duo: Combined GPS/GLONASS Receivers in Urban Environments”, GPS World, January, 2011

(10) Xie, P., Petovello, M. G., and Basnayake, C., “Multipath Signal Assessment in the High Sensitivity Receivers for Vehicular Applications”, Proceedings of ION/GNSS 2011, September 18-22, Portland, OR

(11) Hannah, B., “Modelling and Simulation of GPS Multipath Propagation”, PhD Thesis, Queensland University of Technology, 2001

Author

Peng Xie co-founded OpenLoopNav Inc. Before that, he worked for Samsung and Intel for GNSS receiver design. He received B.S. and M.S. degrees in aerospace engineering from the Beijing Institute of Technology, China, and a Ph.D. from the Department of Geomatics Engineering, University of Calgary, Canada. His current interests include GNSS software receiver design, multipath mitigation and inertial navigation.

The post Multipath Mitigation appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

A set-based urban positioning paradigm.

DANIEL NEAMATI, SRIRAMYA BHAMIDIPATI, 
GRACE GAO, STANFORD UNIVERSITY

3D mapping aided GNSS localization provides state-of-the-art urban positioning by leveraging 3D building maps to account for reduced satellite visibility. Shadow matching is at the core of 3D mapping aided GNSS whereby the user matches the signal degradation at the receiver to the predictions from a 3D building map. Unfortunately, symmetries in building geometry yield multiple regions where the user could be located. With set-based techniques, we can fully account for these ambiguous regions and process the GNSS pseudorange information over each region individually to improve localization.

GNSS Shadows in Urban Environments

City dwellers and urban autonomous systems rely on the Global Navigation Satellite System (GNSS) to provide absolute location services. However, urban infrastructure often degrades standalone GNSS systems [1,2], thereby preventing reliable positioning, navigation and timing. Buildings block, diffract and reflect the line-of-sight (LOS) GNSS signals, thus inducing non-line-of-sight (NLOS) and multipath effects. 3D mapping-aided GNSS (3DMA GNSS) localization has gained traction over the past decade with the increasing availability of high-accuracy 3D city models. Shadow matching is a popular technique for 3DMA GNSS [3,4], among others, such as ray tracing [5,6] and machine learning-based GNSS [7]. Chiefly, the GNSS shadow refers to the areas where city infrastructure blocks direct LOS signals from a GNSS satellite. The user refines the location estimate by determining if the user is inside or outside the GNSS shadow, generally using signal features like signal-to-noise ratio. In this way, the user can turn NLOS and completely blocked signals into valuable information for localization. In past Inside GNSS articles, several authors elaborate on the critical role of shadow matching in 3DMA GNSS [8-11].

While shadow matching improves reliable urban positioning, particularly in the cross-street direction, it also suffers from challenges that restrict its performance. A discussion of the challenges in shadow matching is included in [12], with two two key challenges being a) along-street accuracy is often not reliable and b) multiple positions with large scores yield a multi-modal and ambiguous localization. With a denser urban scene, the location ambiguity often worsens.

We illustrate these challenges in Figure 1, where we extend the common two-dimensional depiction of shadow matching to a slightly larger scene with two streets and different buildings in the foreground and background. The task of shadow matching is to narrow the user location based on the GNSS shadows. For clarity, we only show the shadows of two satellites. The satellite’s shadow is the color-coded region from the building roofs to the ground where the user would receive a highly degraded (i.e., NLOS) GNSS signal or no signal. If the user is on the street outside and has received an NLOS signal or no signal from both satellites, the magenta segments are the only valid user position sets. With only two moderate-elevation satellites, we significantly narrow the user’s location. However, we have not narrowed the along-street dimension (i.e., foreground and background) and we have multiple disjoint valid user sets, so the localization is ambiguous.

We further illustrate these issues of ambiguous localization in Figure 2 as a top view of the scene. The azimuth distribution of satellites is often helpful to localization, especially in cities where building height is variable and buildings have gaps between structures. In Figure 2, the satellites are roughly 140 degrees separated in azimuth. From the top view, we can fully detail the valid user set as a 2D polygon (magenta). In this example scene, there are five disjoint sets for the user position sets that match the user being on the street outside and having received an NLOS signal or no signal from both satellites. We could further reduce the valid user set into smaller sets with additional satellites. However, we are often left with multiple disjoint regions that match the available GNSS shadow information, especially in the along street direction [13].

One possible strategy for improving shadow matching’s along-street accuracy and reducing multi-modal ambiguities in localization is to fuse shadow matching with GNSS pseudoranges. Several authors pioneered this work in improving the urban position accuracy via weighted integration of shadow matching position solutions with that of likelihood-based 3D-mapping-aided GNSS pseudorange ranging [14-16]. The different integration options are reviewed and analyzed in [16]. These methods were further developed into a multi-epoch grid filtering framework in [17,18], which demonstrated improved along-street and cross-street accuracy.

These works built upon prior shadow matching filtering work, such as [19], that combined shadow matching in particle filter and Kalman filter frameworks to resolve multiple modalities over time.

The Set-Based Positioning Paradigm

While past 3DMA GNSS techniques have been successful, they rely on formulating shadow matching in a grid-based manner, which may not be as suitable as a set-valued approach for some users. As illustrated in Figures 1 and 2, shadow matching can be geometrically posed in the following set-based terms: the user is either in the shadow (which is a 2D set or polygon) or outside the shadow (which is the complement set). Set-based formulations conveniently circumvent the need for a grid of position candidates or discretization of elevation and azimuth angles, which is present throughout the aforementioned grid-based works.

Early works in set-based shadow matching [3, 20, 21] struggled to match the computational efficiency of grid-based shadow matching and handled the buildings on a surface-by-surface basis with raster-based techniques that were difficult to scale to dense scenes. In our prior work [13,22], we independently derived set-based shadow matching and designed a novel set-based technique known as Zonotope Shadow Matching (ZSM). Unlike [20, 21], ZSM formulates the entirety of shadow matching with set-based objects. That is, even the buildings are stored as three-dimensional sets. Using the mathematics of constrained zonotopes, we efficiently compute the shadows online using fast vector concatenation operations.

ZSM then iteratively performs set intersection and subtraction to refine a set-based Area of Interest (AOI), i.e., the extent of the depicted ground (black) in Figures 1 and 2. A more complete discussion of the mathematics is included in [13, 22, 23]. Importantly, ZSM may be the algorithm of choice for users who require changes in scales (e.g., from the large scale of a coarse estimate to the small scale of a refined estimate), both online and offline computational efficiency, set-based continuum localization for downstream processing, and complete shadows in a minimal memory representation. In this way, we endeavor to introduce a new set-based paradigm to shadow matching.

To incorporate the GNSS pseudorange information, we leverage the set-based framework from ZSM to form a set-based method to process the GNSS pseudoranges in our recent work [24]. We then develop an iterative set-based filter that exploits the set-based form of the GNSS pseudoranges.

First, we propose Satellite-Pseudorange Consistency (SPC) objects that use the satellite position and pseudorange measurement to transfer set-based information in the two-dimensional receiver position domain among the satellites. The multipath and NLOS effects are notoriously difficult to efficiently model in urban settings [5,6]. The set-based SPC representation allows an efficient, compact, conservative representation of the uncertainty bounds without performing computationally expensive ray tracing. In essence, we trade off the precision of ray-tracing techniques for a more tractable and conservative set-based approach. As discussed in prior works [12, 15, 16], GNSS pseudoranges are most informative in the along-street direction. Second, we fuse a recent history of pseudorange measurements via an iterative filter. Our strategy shares some conceptual similarities with the hypothesis-domain integration in [15, 16] in that we integrate the information at the hypotheses level. However, we diverge significantly from these works with (1) using set-based projections rather than scoring over a grid, (2) explicitly reducing the mode ambiguity, (3) exploiting the slow shadow change compared to the pseudorange variability, and (4) not requiring an NLOS error distribution (e.g., past works assume skewed normal innovation vectors [16]). Our new method directly addresses the challenges of along-street inaccuracy and multi-modal ambiguity reduction identified by [12]. But, our method also significantly relaxes the requirements on shadow matching initialization, model discretization, and uncertainty quantification, all of which [12] considers important advances to shadow matching robustness for reliable urban GNSS localization.

Set-Based GNSS Pseudorange Processing

Unlike other works in urban localization, we leverage the four-dimensional conic geometry of the pseudorange measurement model to handle the pseudorange measurements in a fully set-based framework. Explorations of the four-dimensional conic geometry are largely constrained to the analytical GPS literature [25-28]. We combine the clock bias, environment bias (e.g., multipath) and additional noise into a single term called the range offset. The satellite-dependent range offset is approximately shared across satellites when the receiver clock bias dominates the range offset, the signal is LOS, or when the biases are similarly positively correlated across satellites. The receiver state reflects both the receiver position and overall range offset.

In shadow matching, we implicitly assume the shadows are cast onto a ground plane, as in Figure 1. As noted in [14], terrain aiding significantly improves urban localization, especially when processing pseudoranges. In terrain-aiding, we constrain the receiver state with information on the terrain model and a rough estimate of the receiver height.

This terrain information restricts the user state, thereby yielding a hyperboloid in the three dimensions. This represents all the receiver states consistent with the satellite position, corrected pseudorange and terrain. We call this the Satellite-Pseudorange Consistency (SPC).

The satellite elevation describes the trade between the horizontal plane of the ground versus the vertical. So, the shape of the hyperboloid changes with the satellite elevation angle where the slopes of the hyperboloid near the peak are more shallow as the elevation increases. We provide an example SPC hyperboloid in Figure 3. The zero range offset plane in Figure 3a illustrates the circle in horizontal position space (x, y) consistent with no range offset between the true range and corrected pseudorange.

At the scales of city blocks, the surface is nearly perfectly planar, even for satellites at high elevation angles (Figure 3b). So, we can linearize the hyperboloid about the center of the AOI. We denote the linearization of the SPC hyperboloid as the SPC plane. More mathematical details are provided in [24].

Iterative Set-Based Filter

We design a filtering framework that iteratively combines the information from ZSM and the SPC planes. We summarize the core intuition of the filter in Figure 4. We start with a uniform prior belief over the disjoint sets (magenta in Figure 4, matching Figure 1). First, we construct the SPC planes for each GNSS signal while fixing the user operating height. In blue and orange, we include the SPC planes of satellites 1 and 2 in the foreground of Figure 1. We include the SPC planes of three additional LOS satellites (green) for filter illustration purposes. A base WLS solution with terrain-aiding would find the point with minimum distance to the SPC planes, which can be far from the user position in urban settings. In contrast, we leverage the disjoint sets from ZSM to reduce the locations that the user can be. We form a mixture model to fuse information across satellites in the range offset domain. We weigh the satellites with the probability that the satellite is a LOS satellite. We then find the disjoint set where the fused information is most consistent to determine the more likely option of the disjoint sets from ZSM. When a few LOS satellites are present, this is largely where the LOS satellite SPC planes are nearest each other. In Figure 4, the left magenta set is more likely than the right magenta set because the LOS satellite SPC planes are closer and more overlapping in the range offset domain. From there, we iterate over multiple timesteps to better identify that the left set is the correct set for the user location. More mathematical details are provided in [24].

Performance with Real-World Data

We test the impact of both parts of our approach from [24]: (1) the LOS-weighted set-based SPC projections and (2) the iterative set-based filter. We assess the first part by comparing the SPC projections to shadow matching alone (i.e., ZSM). We further assess the first part by comparing the LOS-weighting in the mixture model to the unweighted mixture model. We test the second part of the approach by comparing a single-step filter to the iterative filter. We validate the filter performance with both a small and a large AOI.

1. Experiment platform and LOS classifier

We collected static-user data with the GNSS Logger App at 1 Hz on a Pixel 3 phone in the Financial District of San Francisco. The user is at the curb on Fremont Street north of Mission Street and outside the Salesforce West building.

The user environment is a significant urban canyon with three prominent glass-facade buildings and two prominent buildings with mixed concrete-glass facades, as illustrated in Figure 5. For ease of processing the signals from the GNSS Logger App, we only use GPS L1 signals though the method herein discussed can be extended to multi-constellation and multi-frequency in future works. We use the same 150 s timeseries throughout the analyses.

We trained a probabilistic LOS Classifier in MATLAB using the TU Chemnitz smartLoc dataset [29]. We trained on the Frankfurt Main Tower, Frankfurt West End Tower and Berlin Potsdamer Platz sections. We tested on the Berlin Gendarmenmarkt section. We use logistic regression and only input the C/N₀ data in the classifier. The final classifier has a 0.5-probability decision boundary at 34.5 dB-Hz between NLOS (negative class) and LOS (positive class). On the test set, we achieve a true positive rate of 69.8% and a true negative rate of 88.3%. Although the smartLoc dataset uses a ublox receiver, the logistic regression classifier generalizes well to the Pixel 3 phone. Further fine-tuning to adapt to the Pixel 3 phone would strictly improve classification but is outside the scope of this article.

2. Set-based shadow matching results

We use the ZSM algorithm detailed in [13, 22] to perform set-based shadow matching. Figure 6 illustrates the results of ZSM for a small AOI (120 m × 120 m in along and cross street directions) and a large AOI (300 m × 300 m in along and cross street directions). We observe two disjoint sets (i.e., a bimodal distribution) in the small AOI case with mode 2 (orange) as the correct mode. We incur six disjoint sets (i.e., six modes in the distribution) when we expand to a large AOI. Standard weighted least squares (WLS) incorrectly predicts mode 1 is the correct mode based on proximity throughout most of the experiment. For the large AOI, WLS incorrectly predicts modes 3, 4 and 5 at select time instances.

3. Set-based location ambiguity reduction

We first analyze how well the method components reduce the localization ambiguity for the case with two disjoint sets (Table 1).

The top performing combination is the proposed method (bottom right corner of Table 1) that starts with ZSM, uses the SPC projections, weights the measurements with the LOS classifier, and iteratively processes the pseudoranges over time.

We correctly arrive at the set with the user’s location in all timesteps with our proposed method for this data set. We also demonstrate how the iterative filter, SPC projections and LOS classifier work together to achieve the sought performance. First, the pseudorange information embedded in the SPC projections is critical in selecting the correct disjoint set. We identify the correct disjoint set in 78% of timesteps (from 0% in the uniform prior with ZSM alone) simply by including the SPC projections, even with a single-step filter. We improve to 96% when querying the LOS classifier to weigh the measurements. If we use an iterative design instead of the LOS classifier, we improve to 99%. Both these options improve the filter by rejecting the spurious NLOS and multipath-ridden outliers either by classification in the former case or by the temporal dispersion of the error in the latter case. We can reap the benefits of both options when we combine them in our proposed method because they work via different mechanisms. With the computational efficiency of the set-based method, the filter calculates the filter updates at roughly 3.7 ms per timestep and is fast enough for real-time operations. 

The second case with six disjoint sets is more difficult for the filter as it must reject five incorrect sets in the face of conservative approximations in the SPC projections. Still, we see the filter identifies the correct set in all timesteps for this dataset (Table 2). Indeed, we arrive at similar results to the case with two disjoint sets. As before, the SPC projections are the most significant improvement as we move from an inability to identify the correct set in ZSM alone to identifying the correct set in 76% of the timesteps with the GNSS pseudorange. However, to achieve the sought performance of 100%, we require input from the LOS classifier and the iterative filter design. 

The time to evaluate 150 timesteps only increases by roughly 200-300 ms (equivalently, 1-2 ms more per timestep) over the case with two disjoint sets. The method easily scales to larger AOIs with more multi-modal distributions.

Conclusion

We presented a new set-based paradigm for urban positioning. Our method reformulates past 3D mapping-aided techniques with computationally efficient set-based operations. In set-based shadow matching, we can fully represent the GNSS shadows without any discretization to better capture the shadow geometry. However, we retain similar issues of ambiguous locations where shadow matching produces multiple disjoint sets where the user could be located. To remedy this, we presented a fully set-based method to reduce location ambiguities in set-based shadow matching. Our proposed method had two key components: (1) processing GNSS shadows in a way conducive to set-based operations; and (2) iteratively filtering the pseudorange information via set-based operations to identify the most likely disjoint set from shadow matching. We validated our approach on smartphone data collected in the dense urban Financial District of San Francisco. We demonstrated both parts of the ambiguity reduction approach are critical to identifying the disjoint set that correctly matched the user location.

Our method is highly computationally efficient, and we can run the filter in roughly 3.7-5.4 ms per timestep depending on the number of disjoint sets. Given the 1 Hz data collection frequency in smartphones, this computational load is suitable for real-time operations. Our ongoing work includes leveraging higher-fidelity maps, quantifying the impact of classification or map uncertainty on the user’s positioning solution, and studying our set-based urban positioning paradigm in more diverse urban settings. 

Acknowledgements

This material is based upon work supported by the National Science Foundation under Grant No. DGE-1656518. We would like to thank Shubh Gupta for reviewing portions of this article. Lastly, we would like to thank the Google Android Location team for free and open-source data processing tools for smartphone GNSS data.

References 

(1) Hsu, L.-T. (2017). Analysis and modeling GPS NLOS effect in highly urbanized area. GPS Solutions, 22(1):7.

(2) Zhu, N., Marais, J., Betaille, D., and Berbineau, M. (2018). GNSS position integrity in urban environments: A review of literature. IEEE Transactions on Intelligent Transportation Systems, 19(9):2762–2778.

(3) Groves, P. D. (2011). Shadow Matching: A new gnss positioning technique for urban canyons. NAVIGATION, 64(3):417–430. Groves, P. D. and Adjrad, M. (2019). Performance assessment of 3D-mapping–aided GNSS part 1: Algorithms, user equipment, and review. NAVIGATION, 66(2):341–362.

(4) Wang, L., Groves, P. D., and Ziebart, M. K. (2015). Smartphone shadow matching for better cross-street gnss positioning in urban environments. NAVIGATION, 68(3):411–433.

(5) Miura, S., Hsu, L.-T., Chen, F., and Kamijo, S. (2015). GPS error correction with pseudorange evaluation using three-dimensional maps. IEEE Transactions on Intelligent Transportation Systems, 16(6):3104–3115.

(6) Suzuki, T. (2016). Integration of GNSS positioning and 3D map using particle filter. In Proceedings of the 29th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2016), pages 1296–1304.

(7) Van Diggelen, F. and Wang, J. (2020). Improving urban gps accuracy for your app.

(8) Wang, L., Groves, P., and Ziebart, M. (2013). Urban Positioning on a Smartphone. Inside GNSS, page 44–56.

(9) Irish, A., Iland, D., and Madhow, U. (2015). Urban Localization and 3D Mapping Using GNSS Shadows. Inside GNSS, page 60–66.

(10) Groves, P. (2016). It’s Time for 3D Mapping–Aided GNSS. Inside GNSS – Global Navigation Satellite Systems Engineering, Policy, and Design, page 50–56.

(11) Groves, P., Zhong, Q., Faragher, R., and Esteves, P. (2021). Supercorrelation Plus 3D Mapping-Aided GNSS. Inside GNSS.

(12) Groves, P. D., Wang, L., Adjrad, M., and Ellul, C. (2015). GNSS Shadow Matching: The Challenges Ahead. In Proceedings of the 28th International Technical Meeting of The Satellite Division of the Institute of Navigation (ION GNSS+ 2015), pages 2421–2443.

(13) Bhamidipati, S., Kousik, S., and Gao, G. (2022). Set-valued shadow matching using zonotopes for 3d-map-aided GNSS localization. NAVIGATION: Journal of the Institute of Navigation, 69(4).

(14) Adjrad, M. and Groves, P. D. (2017). Enhancing Least Squares GNSS Positioning with 3D Mapping without Accurate Prior Knowledge. NAVIGATION, 64(1):75–91.

(15) Adjrad, M. and Groves, P. D. (2018). Intelligent Urban Positioning: Integration of Shadow Matching with 3D-Mapping-Aided GNSS Ranging. NAVIGATION, 71(1):1–20.

(16) Adjrad, M., Groves, P. D., Quick, J. C., and Ellul, C. (2019). Performance assessment of 3D-mapping-aided GNSS part 2: Environment and Mapping. NAVIGATION, 66(2):363–383.

(17) Zhong, Q. and Groves, P. D. (2021). Multi-Epoch 3D-Mapping-Aided Positioning using Bayesian Filtering Techniques. In Proceedings of the 34th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2021), pages 195–225.

(18) Zhong, Q. and Groves, P. D. (2022). Multi-Epoch 3D-Mapping-Aided Positioning using Bayesian Filtering Techniques. NAVIGATION, 69(2).

(19) Wang, L. (2014). Kinematic GNSS shadow matching using a particle filter. In Proceedings of the 27th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2014), pages 1907–1919.

(20) Yozevitch, R., Ben Moshe, B., and Levy, H. (2012). Breaking the 1 meter accuracy bound in commercial GNSS devices. In 2012 IEEE 27th Convention of Electrical and Electronics Engineers in Israel, pages 1–5.

(21) Yozevitch, R. and Ben Moshe, B. (2015). A Robust Shadow Matching Algorithm for GNSS Positioning. NAVIGATION, 62(2):95–109.

(22) Bhamidipati, S., Kousik, S., and Gao, G. (2021). Set-Valued Shadow Matching using Zonotopes. In Proceedings of the 34th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2021), pages 2373–2390.

(23) Scott, J. K., Raimondo, D. M., Marseglia, G. R., and Braatz, R. D. (2016). Constrained zonotopes: A new tool for set-based estimation and fault detection. Automatica, 69:126–136.

(24) Neamati, D., Bhamidipati, S., and Gao, G. (2022). Set-based ambiguity reduction in shadow matching with iterative GNSS pseudoranges. In Proceedings of the 35th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2022), pages 1093 – 1107.

(25) Bancroft, S. (1985). An Algebraic Solution of the GPS Equations. IEEE Transactions on Aerospace and Electronic Systems, AES-21(1):56–59.

(26) Abel, J. and Chaffee, J. (1991). Existence and uniqueness of GPS solutions. IEEE Transactions on Aerospace and Electronic Systems, 27(6):952–956.

(27) Chaffee, J. and Abel, J. (1994). On the exact solutions of pseudorange equations. IEEE Transactions on Aerospace and Electronic Systems, 30(4):1021–1030.

(28) Grafarend, E. W. and Shan, J. (2002). GPS Solutions: Closed Forms, Critical and Special Configurations of P4P. GPS Solutions, 5(3):29–41.

(29) Reisdorf, P., Pfeifer, T., Breßler, J., Bauer, S., Weissig, P., Lange, S., Wanielik, G., and Protzel, P. (2016). The problem of comparable GNSS results – an approach for a uniform dataset with low-cost and reference data. In Ullmann, M. and El-Khatib, K., editors, The Fifth International Conference on Advances in Vehicular Systems, Technologies and Applications, volume 5, page 8. ISSN: 2327-2058.

Authors

Daniel Neamati is a Ph.D. student in the Department of Aeronautics and Astronautics at Stanford University. He received his bachelor’s degree in Mechanical Engineering, with a minor in Planetary Science, from the California Institute of Technology. His research interests include urban GNSS, geospatial information, autonomous decision-making and risk-aware localization.

Sriramya Bhamidipati is a robotics technologist at the Jet Propulsion Laboratory (JPL). Prior to JPL, she was a postdoctoral scholar in Aeronautics and Astronautics at Stanford University. She received her Ph.D. in Aerospace Engineering at the University of Illinois, Urbana-Champaign in 2021, where she also received her M.S. in 2017. She obtained her B.Tech. in Aerospace from the Indian Institute of Technology, Bombay, in 2015. Her research interests include space robotics, GPS, artificial intelligence and unmanned aerial systems.

Grace Gao is an assistant professor in the Department of Aeronautics and Astronautics at Stanford University. Before joining Stanford University, she was an assistant professor at the University of Illinois at Urbana-Champaign. She obtained her Ph.D. at Stanford University. Her research is on robust and secure positioning, navigation, and timing with applications to manned and unmanned aerial vehicles, autonomous driving cars, as well as space robotics.

The post No Signal is also a Signal appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

Identifying seismic signals in GNSS reference stations using machine learning.

TIM DITTMANN, UNIVERSITY OF COLORADO, BOULDER/EARTHSCOPE

JADE MORTON, UNIVERSITY OF COLORADO, BOULDER

Continuous GNSS reference stations represent stable benchmarks for unsung but critical roles in the broader infrastructure: defining reference frames and providing relative corrections, to name a few. But, what if the stable reference station is shaking? A large earthquake will release sufficient energy to permanently deform the earth and vibrate its crust and a coupled GNSS reference antenna¹. Relatively weaker seismic signals at or below the perceived GNSS noise floor still can be problematic for reference products. However, these GNSS seismic ground motions identified amongst GNSS ambient noise are valuable records for seismic monitoring and research. 

In this article, we provide some of our motivation with respect to the scientific utility of ground motion observations, the benefits of using GNSS as a source of these measurements, and the current role of GNSS in seismic monitoring. We then present our work selecting an optimal processing method to pair with a machine learning algorithm. This approach builds on existing stand-alone GNSS seismic awareness to enhance GNSS’ contribution to seismic hazard operations.

Informed Infrastructure Seismic Preparation

Data-driven research allows seismologists to use extensive data archives to account for the complexity of geophysical sources and signal paths of past, present and future events. Catalogs of historical earthquake ground motion data inform models of rupture and energy propagation for informed infrastructure seismic preparation. Real-time ground motion data enable a class of warning called earthquake early warning (EEW) [1]. A successful EEW system detects the earthquake and decides its extent AFTER the earthquake rupture to then alert a population BEFORE peak ground shaking travels to a populated area. This provides those in the impacted area maybe tens of seconds warning to take life-saving actions, whether that’s to duck and cover or to stop a train or medical procedure. Finally, near real-time ground motion data informs maps of shaking intensity for targeted post-earthquake response, and are appended to the historical catalogs as the newest data point for improved preparedness.

These ground motion measurements are the lens into the earthquake system and are traditionally sourced from dedicated, long-standing inertial seismic monitoring infrastructure. The use of higher rate GNSS for seismology, or GNSS seismology [2], was born out of the precision achieved through the seminal engineering of GPS/GNSS and progressed over the last two decades of GNSS seismic research.

Two reasons for including GNSS as a source of seismic observations emphasized in our analysis are:

Increased spatial availability: Existing inertial and geodetic networks were largely built and continue to operate independently. Inclusion of both sensor types increases the density of ground motion observations. Such a densification is particularly valuable in relatively sparser regions [3], such as Alaska, but also adds redundancy and resilience to all existing overlapped networks. 

Dynamic range: Inertial instruments are engineered with specific signal spectral characteristics of interest. As a result, inertial instruments are orders of magnitude more sensitive to weaker signals, including p-waves, the earliest smaller amplitude waves of earthquakes, and surface waves from events halfway around the globe. However, seismologists identified that in the nearfield of the largest events (M7.0+) that information encoded in the slower, longer period, large amplitude displacement signals is required to differentiate the magnitudes of these largest events. Traditional inertial sensors struggle to capture this information due to instrumental reasons. GNSS, with no geophysical upper bound, readily provides either direct or single integration large displacement or velocity measurements necessary for this magnitude differentiation at frequencies down to their permanent offsets. 

One important distinction: In this article, we discuss GNSS seismology and present the complementary nature of inertial and geodetic sensors as stand-alone instruments, as this is currently the primary global infrastructure status quo. However, another closely related area of development and promise is seismogeodesy, or tight local integration of these sensors into a single measurement [4].

The USGS ShakeAlert, the operational EEW system in the United States, ingests GNSS data from the western United States geodetic reference networks through a multi-agency and university collaboration to complement inertial data ingestion. Residents of the western U.S. will benefit from this current culmination of nearly two decades of multi-national GNSS seismology research and engineering. GNSS displacements have been included in USGS post-process fault models [5], but not yet included in shaking intensity products or operational ground motion models.

The potential measurement range of GNSS seismology has not yet been realized operationally in part because of inherent GNSS noise characteristics. GNSS ambient position noise is predominantly the aggregate of timing effects of GNSS radio signal propagation through the atmosphere, satellite and receiver oscillators and the antenna radio frequency environment. Each are location- and time-varying influences, and distinct from the zero-baseline inertial sensor noise seismologists are most accustomed to. Current methods for discriminating signal from noise adopt variations on low-pass filters or static or dynamic thresholds from seismology. To gain the desired sensitivity to signals, high-levels of false alerts from these methods are mitigated through correlating with the inertial system as well as additional GNSS locations. This mitigation adds points of failure and latency and reduces the overall range of valid measurements included. The performance opportunity for improved seismic monitoring and EEW is to rapidly include additional, potentially spatially diverse and unsaturated, ground motion signals from GNSS sources with minimal delay by accommodating the higher dimensionality of GNSS noise.

Comparing Geodetic Processing Methods

To address the presence of seismic signals in GNSS data, we began with an evaluation of two geodetic processing algorithms [6]. Currently, most operational systems and research approaches ingest one of the various methods of precise point positioning (PPP-AR) to make continuous estimates of antenna positions in a global reference frame. These PPP algorithms accomplish this precision at approximately sub-centimeter level using sophisticated error corrections models from multiple sources to estimate carrier phase ambiguities. GNSS seismology requires only relative topocentric motion; consequently, PPP absolute estimates are flattened to relative east, north and up components from a reference position. Time differenced carrier phase processing (TDCP) is a lightweight processing technique first applied to seismic applications by [7]. TDCP single differences epoch-wise carrier phase measurements remove correlated error sources (e.g. troposphere). After removing the satellite velocity, a broadcast ephemeris is acceptable for this, and a least squares system of equations of all observed satellites resolves a topocentric antenna velocity vector and clock drift estimate.

We compared the relative signal to noise of PPP to TDCP to determine our processing method. For our noise estimates, we assembled a dataset of event-free 1 Hz GNSS observational data tracked by multiple receiver types, using a variety of antennas in diverse RF environments, across a hemispheric scale to account for a wide range of noise sources (Figure 3). For PPP processing, we used the UNAVCO/EarthScope PPP solutions from the Trimble RTX software [8]. For TDCP processing, we used the open-source python package SNIVEL [9], which uses GPS only, broadcast ephemeris and the narrow-lane L1/L2 carrier phase combination. From the event-free processed time-series, we estimated a stochastic noise for each station-processing method pair without cleaning or filtering the data. We used these thresholds to establish a statistical noise threshold distribution across this network wide dataset to represent the ambient noise distributions.

For our reference signals, we used empirical scaling laws that relate peak dynamics, earthquake magnitude and radius from the hypocenter. These scaling laws [10, 11] are derived from existing earthquake catalogs and useful for rapid magnitude estimation; the PPP-derived peak ground displacement (PGD) scaling law is part of the current ShakeAlert geodetic contribution. We estimated a signal-to-noise (SNR) metric using PGD or peak ground velocity (PGV) derived from respective scaling laws as our signal reference related to respective ambient noise levels. This SNR metric was estimated over a range of radii from a range of earthquake magnitudes. We found TDCP is more likely than PPP to detect the intermediate magnitude earthquakes (Figure 4) and has additional benefits of being a computational light-weight, open source processing method that doesn’t require external corrections for ephemeris or error source models.

Identifying Seismic Events in GNSS Timeseries with Machine Learning

The results of our ambient processing method comparison indicated TDCP offered lightweight geodetic processing with increased sensitivity, yet still demonstrated unacceptable operational false alarm rates from our statistical threshold. The complexity of GNSS noise coupled with the variability in seismic signals encouraged us to look to an alternative detection approach. Machine learning (ML) is now an ubiquitous tool in data science. Earth scientists have leveraged algorithms developed for natural language processing or image classification and applied them to a range of challenging problems [12] difficult to represent in physics-based models.

We set up a data-driven ML pipeline to train, validate and test a binary classification machine learning model [13]. The foundation of our data-driven experiment is a catalog of 1,706 5Hz TDCP velocity waveforms processed from the UNAVCO/EarthScope geodetic archive concurrent with 80 earthquakes ranging from magnitude of 4.9 to 8.2. An event-free 5Hz TDCP dataset from 30 minute windows prior to the events was included to ensure sufficient noise samples in training and testing for model generalization of these imbalanced datasets. We used 5Hz data to boost signal energy and reduce the likelihood of aliasing, and set a radius of sensitivity for each event as a function of magnitude given our previous sensitivity analysis.

Feature engineering in ML is the process of applying relevant domain knowledge to the ML model for successful classification. We evaluated several feature engineering strategies: the most effective strategy consisted of a combination of time- and lower frequency-domain features (1-30s period) extracted from overlapping 30 second windows. The three topocentric components’ features were labeled through visual inspection and concatenated into a single binary sample and label for each timestamp.

We chose a random forest classifier as our ML algorithm and adopted a nested cross validation technique in our classification training and testing  (Figure 5). This validation strategy allowed us to make training/testing splits of our data on the 80 discrete earthquakes, and evaluate our model’s performance on unseen events in training. We optimized the model on a balance of sensitivity scores and false positives using its F-1 score. A traditional accuracy metric on highly imbalanced classification data, such as our earthquake catalog, is typically not descriptive of performance (e.g., for events happening <1% of the time, you can miss 100% of the events and still be >99% accurate).

The random forest classifier achieved a 90% true positive rate of the station-event pairs (Figure 6) across the entire catalog. The stand-alone classifier substantially outperformed the existing threshold and filtering (e.g. short term average over long term average, STA/LTA) detection methods as shown in Figure 7. These performance results from the classifier’s combination of time- and frequency-domain features into its decision criteria could readily improve GNSS contribution to operational seismic monitoring and ground motion catalogs. Additional investigations in deeper learning models will likely enable researchers to ask more sophisticated questions.

Finally, we tested the timing of the classifier when run once per second on the 5Hz samples of test data not used in training. We found the classifier typically had its first detection approximately at or immediately after the anticipated seismic secondary wave arrival (Figure 8). This result explains our model, or alignment of our results with our domain knowledge that explains the model’s performance. The model did not detect the weaker seismic primary wave arrivals, but instead identified the larger, lower frequency ground motions of the seismic secondary and surface waves. This result also offers implications for GNSS and inertial complimentary hazard monitoring, particularly EEW when timing and accuracy are critical.

Conclusions

Ground motion observations are the data currency of earthquake hazard preparation, monitoring and research. Continuous high-rate GNSS reference stations offer an alternative source that expands the dynamic range of inertial-based ground motion measurements in the nearfield of the largest, most devastating earthquakes and spatially complements existing inertial infrastructure. Complex GNSS noise signatures have bounded operational incorporation of GNSS in these hazard systems. However, alternative, lightweight processing (TDCP) paired with machine learning (random forest classifier) offers enhanced confidence in signal from noise discrimination to confidently include these ground motion measurements in operational systems with minimal false alerting and without external corrections services. The global proliferation of higher-rate GNSS reference stations to support a variety of disparate position, navigation and timing applications could all become medium to large earthquake seismometers, alerting reference station users in addition to contributing to the global seismic monitoring systems. Furthermore, embedding TDCP processing coupled with ML at high rates (>=5Hz) at the network edge will enhance the next generation of geodetic sensor networks to stream higher rate velocities for seismic monitoring or archive denser raw observables for addressing future seismic research objectives. 

Acknowledgments

We would like to thank Yuinxang (Leo) Liu, Kathleen Hodgkinson, Brendan Crowell, David Mencin and Glen Mattioli. We acknowledge the open geodetic data available from the National Science Foundation GAGE facility operated by EarthScope and the open-source software used for GNSS velocity processing and their analysis, including GNSS velocity processing and machine learning libraries.

References

[1] R. Allen and D. Melgar, “Earthquake Early Warning: Advances, Scientific Challenges, and Societal Needs,” Annual Review of Earth and Planetary Sciences, 2019. 

[2] K. Larson, “GPS seismology,” Journal of Geodesy, 2008. 

[3] R. Grapenthin, M. West and J. Freymueller, “The Utility of GNSS for Earthquake Early Warning in Regions with Sparse Seismic Networks,” Bulletin of the Seismological Society of America, 2017.

[4] Goldberg, D. E., and Y. Bock (2017), Self-contained local broadband seismogeodetic early warning system: Detection and location, J. Geophys. Res. Solid Earth, 122, 3197–3220, doi:10.1002/2016JB013766.

[5] D. E. Goldberg, P. Koch, D. Melgar, S. Riquelme and W. L. Yeck, “Beyond the Teleseism: Introducing Regional Seismic and Geodetic Data into Routine USGS Finite‐Fault Modeling,” Seismological Society of America, 2022.

[6] T. Dittmann, K. Hodgkinson, J. Morton, D. Mencin and G. Mattioli, “Comparing Sensitivities of Geodetic Processing Methods for Rapid Earthquake Magnitude Estimation,” Seismological Research Letters, 2022.

[7] G. Colosimo, M. Crespi and A. Mazzoni, “Real‐time GPS seismology with a stand‐alone receiver: A preliminary feasibility demonstration,” Journal of Geophysical Research: Solid Earth, 2011.

[8] R. Leandro, H. Landau, M. Nitsschke and e. al., “RTX positioning: The next generation of cm-accurate real-time GNSS positioning,” Proceedings of the 24th international technical meeting of the satellite division of the Institute of Navigation, 2011.

[9] B. W. Crowell, “Near-field strong ground motions from GPS-derived velocities for 2020 Intermountain Western United States Earthquakes,” Seismological Research Letters, 2021.

[10] D. Melgar, B. Crowell, J. Geng, R. Allen and Y. Bock, “Earthquake magnitude calculation without saturation from the scaling of peak ground displacement,” Geophysical Research Letters, 2015.

[11] R. Fang, J. Zheng, J. Geng, Y. Shu and C. Shi, “Earthquake Magnitude Scaling Using Peak Ground Velocity Derived from High-Rate GNSS Observations,” Seismological Research Letters, 2020.

[12] K. Bergen, P. Johnson, M. V. de Hoop and G. Beroza, “Machine learning for data-driven discovery in solid Earth geoscience,” Science, 2019.

[13] T. Dittmann, Y. Liu, J. Morton and D. Mencin, “Supervised Machine Learning of High Rate GNSS Velocities for Earthquake Strong Motion Signals,” Journal of Geophysical Research: Solid Earth, 2022.

Authors

Tim Dittmann is a data scientist at the EarthScope consortium and doctoral candidate at the Ann and H.J. Smead Aerospace Engineering Sciences at the University of Colorado, Boulder.

Y. Jade Morton is Helen and Hubert Croft Professor and Director of the Colorado Center for Astrodynamics Research in the Ann and H. J. Smead Aerospace Engineering Sciences Department at the University of Colorado Boulder. She received a Ph.D. in Electrical Engineering (EE) from Penn State. She is a member of the U.S. Space-based PNT Advisory Board, a recipient of the AGU SPARC award, the IEEE PLANS Kershner Award, and the Institute of Navigation’s (ION) Burka, Thurlow, Kepler, and distinguished service Awards. Dr. Morton is a Fellow of ION, RIN and the IEEE.

The post Expanding the Role of GNSS in Seismic Monitoring appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

A: While GPS satellite faults have always been infrequent, they have become rarer in the past decade due to the maturity of the GPS constellation and its Operational Control Segment (OCS). Lessons learned from past GPS satellite failures have also aided the development of newer GNSS constellations such as the European Galileo.

Rebecca Wang, Todd Walter, Sam Pullen, Stanford University

The resulting improvements in GNSS performance and reliability mostly go unnoticed by the majority of GNSS users but are real and important.

The civil community that supports GNSS users with demanding integrity requirements (meaning tight limits on the probability of unbounded and undetected errors) needs to understand both the probability and dynamics of different classes of satellite failures to certify that augmentation systems such as SBAS and GBAS and algorithms for unaugmented users adequately protect all user classes and levels of service. Automated methods have been developed to analyze GNSS satellite measurements and navigation data recorded over time by global reference station networks such as the International GNSS Service (IGS) at igs.org. The results of these studies have been used to both better understand rare-event GNSS satellite behavior and confirm that performance standards relied on by civil users, such as those in the GPS Standard Positioning Service Performance Specification (GPS SPS PS) [1], have been and continue to be met with margin (for more on this, see the “GNSS Solutions” article in the May/June 2022 issue of Inside GNSS).

This article summarizes recent results published in [2] showing examples of GPS and Galileo satellite faults and “near faults” over the past six years (2017 to 2022) and into 2023. The focus here is on “integrity faults,” meaning the subset of satellite faults that are not immediately obvious to users and thus could lead to loss of user integrity if not detected and alerted by augmentation system or user monitoring. These are distinct from more-common satellite outages triggered by loss of signal tracking or health parameter flags in navigation data. In this article, integrity faults are often just called “faults” for simplicity.

GNSS Satellite “Integrity Faults” and “Near Faults” 

The definition of a satellite integrity fault for both GPS and Galileo is a range error that exceeds K_F  σ_URA, where σ_URA represents a one-sigma bound on satellite User Range Accuracy provided by the constellation, and K_F is a unitless multiplier. σ_URA is a bound that allows use of the standard Normal (Gaussian) distribution with zero mean and the given standard deviation, thus K_F is based on the standard Normal deviation at the specified integrity probability. For example, the probability of an individual GPS satellite being in an integrity-faulted state should not exceed 10^-5 according to [1]. The standard Normal deviation that exceeds this probability, considering both positive and negative errors, is 4.42, thus K_{F_GPS} = 4.42. The value of this probability for Galileo is 3×10^-5 from [3], thus K_{F_Galileo}= 4.17.

GPS and Galileo are different in that σ_URA for GPS is included within the broadcast navigation data and, for Legacy navigation (LNAV) data, is usually 2.4 meters but occasionally rises to 3.2 meters or higher. Galileo, on the other hand, specifies a fixed value of 6 meters for σ_URA when two frequencies (E1 and E5a) are used to estimate and remove ionospheric delay from the measured pseudoranges [3]. While this may change in the future, the current range error threshold for Galileo integrity faults (6 meters × 4.17 ≅ 25.0 meters) is much looser than the most common GPS value (2.4 meters × 4.42 ≅ 10.6 meters).

In this article, the maximum range error among user locations that can view and track the satellite is evaluated and compared to the fault criteria. The normalized maximum projected error (MPE) is this error divided by the corresponding σ_URA. This normalization (MPE/σ_URA) reduces the fault definition to the comparison between MPE and K_F, and is shown in Figure 1. In addition, because integrity faults are rare, recent GPS “near faults” are also examined. Near faults are defined by normalized MPE exceeding K_NF = 2.5 [2], which is much smaller than K_F but still represents a significant error that should be unlikely (with a probability of no greater than 0.0124 based on the standard Normal distribution).

Recent GPS Service History

Figure 1 shows an overview of the normalized MPE for all GPS satellites reporting healthy status from the beginning of 2017 to the end of 2021 (at 300-second intervals). Because the value of σ_URA in the GPS LNAV message is conservative, it is not surprising that the vast majority of the normalized MPE points in this plot are within –1 and 1. Some points exist between 1 and 2 (on either side of zero), but only four events exceed 2.5 and represent “near faults.” They are highlighted by red boxes in Figure 1 and are listed in Table 1. However, no actual fault events were observed. Three of the four near-fault events occurred in the last year of this period (2021), which might be due to chance but suggests that near faults might be becoming more frequent. Of greater importance is the fact that, since the end of 2021, two faults (not just near faults) have occurred. These are shown in Table 2. The first of these, on October 2, 2022, was the first integrity fault observed on GPS in more than 10 years (since June 17, 2012), while the second integrity fault followed it by less than four months.

Figures 2 and 3 show details of the satellite error behavior on GPS PRN 12 (SVN 58) that led to the fault event on October 2, 2022 [2]. In Figure 2, the definition of a fault based on 4.42 times the broadcast σ_URA is shown by the thick blue lines. A slow clock error ramp (the purple curve in Figure 2 and the second plot in Figure 3) began around 10:20 (UT) and increased over five hours before finally crossing the fault definition at about 15:10. Afterward, the satellite remained in a faulted state for about 45 minutes before the satellite transmits updated navigation data (indicated by an IODC change in the bottom plot of Figure 3) that corrects the error. Note the satellite was never flagged as unhealthy and instead returns to nominal performance when new navigation data is uploaded by the OCS and begins being transmitted. This behavior is not typical, but it was also observed in the two previous GPS satellite faults on April 25, 2010 and June 17, 2012 as reported in [4]. It also appears that an earlier navigation data update tried to correct the problem with limited success: the update at around 11:00 reduced the clock error but did not remove the ramp-wise error growth.

Figures 4 and 5 show details of the satellite error behavior on GPS PRN 1 (SVN 63) that led to the most recent observed fault event on January 25, 2023 [2]. Here, the fault was caused by an obvious (in retrospect) clock error jump occurring at roughly 16:10. The fault lasted for about 185 minutes until roughly 19:15, when the satellite is set unhealthy by an OCS navigation upload with a new IODC. The maximum clock error of about 27 meters significantly exceeded the 10.6-meter fault definition, and the duration of the fault was significantly longer than any of the previous fault durations observed since 2008. Once the satellite was flagged as unhealthy, it remained that way until approximately 15:45 on the following day (January 26, 2023), when another upload returned the satellite to a healthy state.

These two recent GPS satellite faults have similar characteristics to those observed from 2008 to 2012 [4], but like the recent uptick in near-fault events, the occurrence of two faults in relatively rapid succession is concerning. The PRN 12 clock fault on October 2, 2022, had less effect on users because the 4.42×σ_URA fault definition was only slightly exceeded after several hours of slow error growth. However, because it appears that several unscheduled navigation data uploads were made, presumably to attempt to correct the error, it is unclear why the GPS OCS did not flag the satellite as unhealthy sooner before correcting the error. One possibility is that, because the error remained small and well below the fault threshold for some time, there was no immediate perceived need to take decisive action. The most recent fault on January 25, 2023, was likely due to anomalous behavior of the clock on the satellite and could not be foreseen or prevented by OCS action, but the delay of more than 3 hours before flagging the satellite unhealthy seems odd, given the resulting MPE suddenly grew to over 2.5 times the fault definition. Section 3.5.1 of the GPS SPS PS [1] cites a maximum fault state duration of 6 hours, which to our knowledge has never been exceeded. The mean fault state duration is specified as 1 hour, which is still met when averaging across all faults observed since 2008.

Recent Galileo Service History

Because the Galileo satellite constellation is much younger than GPS and only achieved Early Operational Capability (EOC) in late 2016, it would not be surprising if it suffered integrity faults more often than GPS, particularly because GPS itself had a higher rate of faults prior to 2008. While large errors did occur on multiple Galileo satellites in July 2019 due to a ground-system problem, this was not an integrity fault because safety-critical users following the guidance in the Galileo OS SDD [3] would not have used the erroneous signals and data [5].

Figure 6 gives a graphic overview of Galileo satellite behavior from January 2018 to the end of June 2022 [2]. It shows periods of known “unhealthy periods” with blue lines and includes the July 2019 event in which many satellites were simultaneously unusable. Four integrity faults are also shown in Figure 6 and are highlighted by pink boxes. Table 3
indicates the affected satellites and the duration of these fault events. The first two of these faults were on the earlier generation of IOV satellites, while the latter two were on the same FOC satellite (SVN 210). All four of these events had short fault state durations of about 40 minutes or less before the satellites were flagged as (potentially) unhealthy based on the broadcast signal-in-space accuracy (SISA) parameter, which changed from indicating finite values to “No Accuracy Prediction Available” or “NAPA.” Satellites with SISA messages of NAPA are either unhealthy or of “marginal” health according to [3] and should not be used in safety critical applications.

Two of the four Galileo satellite faults in Table 3 are illustrated in detail. Figures 7 and 8 show the fault on October 29, 2019 on PRN 11 [2]. As shown in Figure 7
and the second plot in Figure 8, at 17:55, the clock error began to quickly “ramp off” in an unbounded manner. At roughly 18:05, the MPE exceeded the 25-meter Galileo fault definition, and roughly 40 minutes later, at 18:45, the broadcast SISA changed from 3.12 meters to NAPA. The clock error grew to be about 500 meters at the end of the fault duration (when NAPA was broadcast), after which it grew to be greater than 5,000 meters. Meanwhile, the top plot of Figure 8 shows the along-track and cross-track dimensions of satellite ephemeris error also grew in a ramp-like fashion but stayed at nominal values below ± 2 meters during the interval shown. However, the marginal/unhealthy status persisted for more than one day, and during this period, the radial, along-track, and cross-track errors continued to grow and became significant as well. The satellite was not fully restored to healthy and usable status until almost one month later, on November 28, 2019.

Finally, Figures 9 and 10 show the most recent observed Galileo fault on PRN 1 on April 29, 2022 [2]. This fault also was a rapid ramp-like clock error increase and was similar to the October 2019 fault as well as the September 2021 fault on the same satellite. In this case, the ramp started around 00:58 and exceeded the fault definition at around 01:00. The MPE of this fault settled at about 250 m, which is smaller than the 2019 and 2021 faults but is still very large. The fault duration lasted about 12 minutes until a SISA value of NAPA was broadcast at 01:12. After almost one month, on May 25, 2022, the satellite became healthy and usable again.

GPS and Galileo Satellite Integrity Fault Comparison

Three things stand out from this review of GPS and Galileo satellite faults. The first is the dominant cause of these faults is clock error ramps in which the atomic clock driving the satellite signal grows quickly and unpredictably. This is not new—it has been the most common cause of GPS satellite faults since observations began around 2000. Ephemeris error growth sufficient to exceed the definition of a fault has also occurred but is much less common.

Second, while both GPS and Galileo respond to satellite faults quickly, the mechanism of curing the fault condition differs. Both systems indicate faulty conditions by uploading updated navigation data, but GPS sometimes attempts to correct the errors with updated clock and ephemeris navigation data rather than proceeding immediately to “unhealthy” status, particularly when the error growth is relatively slow and straightforward to correct. Galileo instead appears to almost always respond first by updating the SISA parameter to NAPA and attempting correction and repair later. The fact the Galileo faults described here have rapid clock-driven error growth leading to large MPE values is one reason the first Galileo response is to alert loss of health through NAPA. Fortunately, this response is rapid, as the maximum fault duration observed since 2018 is 40 minutes.

Third, integrity faults on both GPS and Galileo remain quite rare and appear to satisfy the promised state probabilities of below 10^-5 for GPS and 3×10^-5 for Galileo with margin. Estimates of these probabilities based on the years of observations reported here and in [4] are given in [2]. GPS has had two integrity faults in the past six months, which is surprising considering the lack of any such faults between 2012 and 2022 (at least as observed by the methods used here).

Summary

This article provides an overview of recent GNSS satellite integrity faults and shows examples of such faults in both GPS and Galileo. While the specifics differ, the nature of these faults and the means by which GPS and Galileo responds to them are broadly similar. The probabilities of these faults remain below the standards issued by both systems. However, the recent occurrence of two GPS faults after many years without any demonstrates that offline monitoring of GPS, Galileo and other GNSS constellations must continue without pause to provide confidence that these standards remain met in the future. 

Acknowledgments

The authors thank the FAA Satellite Navigation Team for funding this work under Memorandum of Agreement 693KA8-22-N-00015 and for their continuous support over many years.

References

(1) GPS Standard Positioning Service (SPS) Performance Standard (GPS SPS PS), Washington DC, U.S. Dept. of Defense, 5th Edition, April 2020. https://www.gps.gov/technical/ps/2020-SPS-performance-standard.pdf

(2) R. Wang and T. Walter, “Characterization and Comparison of Galileo and GPS Anomalies,” Proceedings of ION 2023 International Technical Meeting (ITM). Long Beach, CA, Jan. 2023. http://web.stanford.edu/group/scpnt/gpslab/pubs/papers/Wang_ION_ITM_2023_GNSS_Faults.pdf 

(3) Galileo–Open Service–Service Definition Document (Galileo OS SDD), European Union Agency for the Space Programme (EUSPA), Issue 1.2, Nov. 2021. https://www.gsc-europa.eu/sites/default/files/sites/all/files/Galileo-OS-SDD_v1.2.pdf

(4) T. Walter and J. Blanch, “Characterization of GNSS Clock and Ephemeris Errors to Support ARAIM,” Proceedings of 2015 ION Pacific PNT Meeting, Honolulu, HI, April 2015. http://web.stanford.edu/group/scpnt/gpslab/pubs/papers/Walter_IONPNT_2015_ARAIM_characterization.pdf

(5) E. Chatre and J. Benedicto, “2019–Galileo Programme Update,” Proceedings of ION GNSS+ 2019. Miami, FL, Sept. 2019. https://doi.org/10.33012/2019.1690

Authors

Rebecca Wang is a graduate student in the GPS Research Laboratory working under the guidance of Professor Todd Walter in the Department of Aeronautics and Astronautics at Stanford University. Prior to joining the lab, Rebecca received her B.S. in Aerospace Engineering at the University of Texas at Austin. Her research interests include multi-GNSS integrity for aviation and high-accuracy navigation. 

Todd Walter is a Research Professor in the Department of Aeronautics and Astronautics at Stanford University. He is also a member of the National Space-Based Positioning, Navigation, and Timing (PNT) Advisory Board. His research focuses on implementing satellite navigation systems for safety-of-life applications. He has received the Institute of Navigation (ION) Thurlow and Kepler awards. He is also a fellow of ION and has served as its president.

The post Q: What are typical features of GNSS satellite faults in recent years? Are there any differences between satellite constellations? appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

A look at the state of 5G NR NTN.

If beyond visual line of sight (BVLOS) operations are the “Holy Grail” of the UAS industry, harnessing the capability of non-terrestrial networks (NTNs), or Space-Based Adaptive Communications Node networks, appears to be the equivalent for a 5G communication system network—particularly in low earth orbit (LEO) satellite communications (SatComs) constellations.

UAS, among many other systems, can play a vital role in developing requirements, which could, in turn, help mitigate current technological limitations for BVLOS and a wide range of other use cases that depend on accurate positioning, navigation and timing (PNT). However, as in so many other areas of emerging technology, challenges in implementation and standardization remain unsolved. Here’s a rundown of the state of play for 5G LEO SatCom networks.

The Basic Plan

To appreciate the game-changing nature of 5G New Radio (NR), you must first have a basic understanding of how traditional mobile telecommunications work.

Telecoms traditionally consist of four key components, according to Ericsson, an information and communication technology (ICT) service provider. Most of us engage directly with user equipment (UE) devices such as smartphones and tablets. The Radio Access Network (RAN) wirelessly connects UEs using radio frequencies (RF). Coordination between various parts of the RAN and the connection to the internet occurs through the core network (CN). Finally, the transport network supplies the connection between the RAN and the CN.

Complex integrated hardware and software enable these functions, even in traditional comms. Baseband equipment that performs all of the signal processing functions required for wireless communications (e.g., for multiple antennas, to detect/correct transmission errors, provide security and manage resources) contains high performance electronics and cutting edge software. Radios ensure signal transmissions travel on the correct bands at the required voltage and actually convert digital information into those signals. Antennas beam out those electric signals into radio waves.

5G, which requires Multiple-Input Multiple-Output (MIMO), adds layers of complexity to this basic telecom system. It requires cross-functional integration, such as integrating radios and baseband hardware and software with antennas. 5G NR RAN (which replaces the Long Term Evolution (LTE) high speed and low latency RAN) and CN software can be deployed and managed on the same infrastructure. 

For this reason, RAN radios (baseband and antenna-integrated) and CN sites depend on software, on each other and on complex code. For maximum coverage, companies have built additional base radio stations, called gNB (Next Generation/gNodeB, which replaces the eNB or eNodeB or Evolved Node B) and deployed AI and ML to orchestrate and balance traffic. 

Why does this matter? 5G’s unified interface enables higher speeds, reduces latency and increases the availability and reliability of connections. 

The Value Proposition

According to Qualcomm, 5G will fuel “massive IoT” and drive global growth. The company’s landmark 5G Economy Study found 5G could potentially enable up to $13.1 trillion worth of goods and services across a diverse group of businesses worldwide by 2035. More than 60 countries have already deployed 5G.

Now add in NTNs, which deliver 5G/NR service via space (satellite) or air (airborne platform) to the 5G mix. This multi-layered network can include SatCom networks, high altitude platform systems (HAPS), UAS and other air-to-ground networks.

According to 3GPP, the NTN 5G value proposition is clear. NTN systems can significantly bolster 5G service continuity where a single or series of combined terrestrial networks cannot, particularly for mobility assets and mission-critical communications. NTN can bridge 5G service coverage gaps where terrestrial networks do not exist or simply do not reach. This includes oceans, deserts, wildernesses and urban areas. Scalability, through NTN’s wide area coverage and ability to multicast, also tops its list of benefits. 

NTN 5G can support a wide range of use cases, including aeronautical and maritime tracking systems. Specifically, Automatic Dependent Surveillance-Broadcast (ADS-B), which is based on the capability of the aircraft to navigate to a destination using GNSS data and barometric altitude, allows for communication with air traffic control, cooperative surveillance, separation and situational awareness. It depends on aircraft navigation system data derived primarily from GNSS signals and then broadcast to aircraft and ground-deployed infrastructure. 

But this infrastructure does not exist in a number of areas, including over oceans and in the Arctic. LEO-based ADS-B receivers could contribute to the ATC relay network. This would result in low latency and secure coverage globally. In the maritime sector, the equivalent tracking system, Automatic Identification System (AIS), also benefits from space-based receivers.

The benefits of NTN 5G extend beyond transportation and more broadly for internet of things (IoT) applications, from surveillance of infrastructure to precision agriculture. 

Standards Moving Forward

The 3rd Generation Partnership Project (3GPP), a global partnership of telecommunications standard development organizations, started working on 5G NR NTN about 5 years ago. Its first study, Release 15 (Rel-15) documented in TR 38.811, targeted deployment scenarios and models that included not only LEO and HAPS, but GEO satellites as well. It addressed issues such as relevant beams, elevation angles, satellite deployment footprint, various NTN terminals and antenna arrays. 

The follow on study, Rel-16, focused on minimum viable architecture, higher layer protocols, and physical layer aspects (TR 38.821). This study concluded that the group’s NR work provided a solid basis to support NTN. It identified additional areas of study including: timing relationships, uplink time and frequency synchronization, and hybrid automatic repeat request (HARQ), a combination of high-rate forward error correction (FEC) and automatic repeat request (ARQ), essential for reliable data transmissions.

Earlier this year, ratified Rel-17 focused on 5G system enhancements. Among other things, Release 17 involves physical layer aspects, protocols, architecture and radio resource management. This study assumes all UEs have GNSS capabilities. 3GPP “froze” (meaning no further functions can be added to the specification) the Protocols for this study in March 2022 and the Protocol Code (OpenAPI) in June. According to 3GPP, the “Release 17 Description; Summary of Rel-17 Work Items” (TR21.917) remains in production. 

In the meantime, 3GPP launched the Rel-18 study. It focuses on 5G Advanced, addressing extraterritorial coverage of satellites and high altitude systems.

Progress continues to move forward on the coding side of the house, courtesy of the OpenAirInterface Software Alliance (OSA). This is significant because many of these systems rely on complex code stacks. The OSA, a French non-profit organization established in 2014 and funded by corporate sponsors, is the home of OpenAirInterface (OAI). OAI, an open software endeavor, has gathered a community of developers from around the world who work together to build wireless cellular RAN and CN technologies. 

The OSA OAI 5G Project Group seeks to develop and deliver a 3GPP compatible 5G gNB RAN software stack under the OAI Public License V1.1. In October, the group provided an OAI codebase status update and development roadmap.

Simultaneously, the organization’s related OAI 5G-LEO extension for 5G satellite links aims to use the OAI as a tool to assist in 5G NTN R&D. This 5G-LEO Project has four main objectives, according to the European Space Agency (ESA):

1. Select a 5G-LEO baseline scenario for 3GPP NR-NTN system deployments to implement and verify with the extended OAI library.

2. Identify fundamental codebase gaps and changes to extend OAI to the 5G-LEO baseline. 

3. Implement required OAI code adaptations for the different layers of the 3GPP protocol stack to support 5G-LEO within Rel-17 and potentially in Rel-18.

4. Set up an end-to-end 5G-LEO demonstrator in the lab for experimental validation of the OAI extension for the 5G-LEO baseline scenario.

This two-phase project, started in December 2021, is in its second phase. It’s focused on implementation, software compliance and demonstration.

Challenges and Stratospheric Possibilities

R&D continues to tackle other challenges that must be mitigated for successful LEO-based 5G NTN. Propagation delays and large Doppler shifts caused by moving cells rank high among them.

Propagation delays result in latency. Depending on the satellite’s altitude, long distances between satellite constellations, ground stations, and user terminals cause time delays in radio wave transmissions. While delays from LEO satellites are much shorter than higher altitude GEO satellites, the constellation and ground station deployments must be larger to cover wider areas. This can increase costs. Groups are exploring workarounds such as sat-to-sat relays, or mesh networks for CN functions, to mitigate these issues.

On the military side, the same mitigation measures used to solve common latency issues in GEO MIL/SATCOM applications can be applied to a LEO MIL/SATCOM architecture, said Jason “JD” Danieli, CEO of Colorado-based Giuseppe Space Enterprises. 

“Modeling, simulation and analysis, as we commonly refer to as MS&A, is an extremely important first step prior to deploying new tech. There are so many variables to consider and account for. A common SW tool, such as MATLAB, is just one of many tools we consider when solutioning,” Danieli said. “We attempt to address issues prior to deployment such as interoperability, performance and resiliency.” 

Research also remains ongoing to address Doppler effects in the LEO orbit. This phenomenon describes the increases (or decreases) in the frequency of sound, light or other waves as the source and observer move toward (or away from) each other. Waves emitted by a source traveling toward an observer get compressed. The LEO satellites and airborne platforms for 5G NTN move very fast while the user terminal remains either stationary or moves slowly. This results in large Doppler shifts experienced by the receiver, leading to communication degradations between transmitters and receivers. This is why 3GPP assumes NTN devices will be equipped with a GNSS chipset to determine position and calculate the needed frequency adjustments. 

But GNSS has its own challenges in terms of vulnerability. Efforts to make PNT more resilient continue to churn…slowly. 

On the bright side, LEO satellites offer several attributes that are attractive to supplement GNSS for positioning and timing. This includes an abundance of signals, favorable geometric configurations, and diverse signal frequencies. 

“With 5G NTN entering the stage, we will have even more satellite signals to consider on top of the 3,000+ LEO satellites whose signals have already shown potential to supplement GNSS,” said StarNav CEO Joshua Morales, who has spent nine years building PNT systems that use cellular and LEO satellite signals as a backup to GPS.

New partnerships have cropped up to tackle these challenges and take advantage of the benefits that NTN has for the future of 5G. Last summer, for example, Ericsson, Qualcomm Technologies, Inc. and French aerospace company Thales announced a partnership for the first testing and validation of 5G NTN. This work aims to validate 3GPP’s assumption that 5G NTN can be supported in a smartphone form factor. Initial tests are taking place in an emulated space environment in France. 

With the possibility of successful 5G NTN just within our grasp, we can no longer just say the sky’s the limit—because the possibilities are truly out of this world. 

The post Washington View: Beam Me Up (And Down), Scotty appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

This is the most efficient way to combine INS with aiding data. Here’s a look at the key principles and benefits.

Inertial navigation enables self-contained navigation in any environment. To reduce drift in INS outputs, sensor-fusion mechanizations use data from various aiding sources (such as GNSS, maps, electro-optical sensors, etc.). A complementary fusion methodology represents the most efficient way to combine INS with aiding data. This column discusses its key principles and main benefits.

Navigation Architecture

Figure 1 shows a high-level diagram of complementary sensor-fusion with inertial navigation.

The algorithm uses differences between aiding measurements and their INS-based predictions as inputs to the complementary estimator. An Extended Kalman Filter (EKF) is the most common form of the complementary estimation, however, other estimation methods (such as particle filters and factor graphs) have been applied as well. 

Example Implementation

To provide insights into complementary filter design aspects, this section considers an example case of loose GNSS/INS integration with the EKF-based fusion. The Kalman filter mechanization is completely defined by (i) state vector X, (ii) state transition matrix F, (iii) process noise covariance Q, (iv) measurement (or observation) vector z, (v) observation matrix H, and (iv) measurement error covariance R. Once these six terms are defined, recursive Kalman filter updates are applied using standard filter equations. The INS error propagation defines the first three terms. The propagation mechanism was considered in a previous Inertialist column (see May/June 2022 issue of Inside GNSS). To complete the integrated system mechanization, we need to define the remaining three terms. 

For loose integration, complementary position observables are formulated as differences between INS and GNSS position solutions (rˆ_INS and rˆ_GNSS):

where δr is the INS position error and ε_r is the GNSS position measurement error vectors, respectively. Note the update rate of INS solution is generally higher than that of GNSS. Hence, the observation update at t_n is applied only if GNSS position is available at that time. Otherwise, the complementary filter relies on prediction with state estimates being assigned their predicted values. 

Additionally, the formulation makes two other simplifications. First, INS and GNSS solutions are assumed to arrive at the exact same time (i.e., stay completely synchronized). Generally, this is not the case and timing adjustments must be made. The adjustment can be performed by propagating INS navigation states to the time of validity of GNSS measurements. Second, the inertial measurement unit (IMU) and GNSS antennas are assumed to be collocated with each other (i.e., their relative lever arm vector is zero). Equation 1 needs to be updated to include a lever-arm term for non-zero lever-arm cases.

The observation matrix H defines how elements of the state vector project into measurement observables. This matrix has a size of K×P, where K is the number of scalar measurements and P is the number of elements in the state vector. For position updates (three position components) and a 15-state INS error model (including position, velocity and attitude errors, and gyro and accelerometer biases), H has a size of 3×15. Position error states directly project into the observation vector, while the contribution of other error states is zero. Hence, the H matrix is formulated as:

where I_(3×3) and 0_(3×3) are the 3×3 unit matrix and zero matrix, respectively. 

Formally, the observation matrix is derived by taking partial derivatives:

where: 

H_k,p is the element of the observation matrix that corresponds to its k^th row and p^th column; 

z_k is the k^th element of the observation vector; and, 

X_p is the pth element of the state vector. 

It can be readily verified that when partial derivatives of the observation vector in Equation 1 are computed with regard to the elements of the state vector, this results in the observation matrix in Equation 2.

By definition, the measurement error covariance R is:

where E[ ] denotes the mean value and ^T is the matrix transpose operator.

The Kalman filter assumes that measurement errors are zero-mean Gaussian. R is a diagonal matrix when errors in different GNSS position components are not correlated with each other:

where σ_x , σ_y and σ_z are standard deviations of x, y and z position errors, respectively. 

This example of loosely coupled GNSS/INS can be extended to other complementary sensor fusion mechanizations such as tightly coupled GNSS/INS and fusion with other aiding data sources (e.g., bearing angles to visual landmarks at known or unknown locations). The INS system propagation remains the same, but the state vector may have to be augmented by error states associated with the aiding sensor (such as GNSS receiver clock error states and misorientation between the IMU and video camera). The complementary observations are formulated in a similar manner, i.e., as differences between actual aiding measurements and their INS predicted values, for example, between measured and predicted GNSS pseudoranges. The observation matrix H is derived by taking partial derivatives of the measurement observables similarly to Equations 2 and 3. Finally, the measurement noise matrix R is formulated based on measurement noise covariances. 

Main Benefits

As compared to the full-state formulation (i.e., direct estimation of the navigation states), the main benefit of a complementary filter for GNSS/INS integration is a significantly simplified modeling of state transition. Inertial errors are propagated over time instead of propagating navigation states themselves. In this case, the process noise covariance matrix, Q, is completely defined by the stability of INS sensor biases, as well as sensor noise characteristics. On the contrary, modeling of actual motion generally needs to accommodate different motion segments (such as a straight flight versus a turn maneuver), which can require ad hoc tuning. An adaptive tuning of the Q matrix may be required to optimize the performance. 

To illustrate this benefit, Figures 2 through 4 show example simulation results. A filtering of noisy GNSS position is considered. The motion trajectory includes two straight segments and a 180 degree turn maneuver. Figures 2 and 3 show outputs of a Kalman filter that does not use the INS and model navigation states instead (more specifically, a constant-velocity motion model is applied). In Figure 2, a smaller value of the system noise matrix (Q-matrix) is used. This allows for an efficient smoothing of GNSS measurement noise. However, during the turn (when non-zero acceleration is introduced and the constant-velocity motion model becomes invalid), position errors increase substantially. Increasing the system noise matrix increases the measurement feedback into the Kalman filter and helps to mitigate the divergence during the turn as shown in Figure 3. However, it comes at a cost of increased estimation noise.

The dynamic modeling/noise smoothing trade-off is eliminated when a complementary GNSS/INS Kalman filter is applied as shown in Figure 4. In this case, the system (i) substantially suppresses the measurement noise (similarly to Figure 2), and (ii) accurately follows the motion dynamic (similarly to Figure 3). 

Another benefit of the complementary estimation is the ability to linearize the state propagation and use computationally efficient linear estimation techniques. Specifically, propagation of navigation states through INS mechanization is a non-linear process. In contrast, time-propagation of inertial errors into navigation outputs can be efficiently linearized. As a result, the complementary filter can benefit from linear filtering techniques (such as an extended Kalman filter) without the need for nonlinear estimation methods such as an unscented Kalman filter and particle filter.

Observability of Error States 

State observability of the complementary sensor-fusion is one of the key aspects influencing system performance. To provide an insight into it, this section considers a two-dimensional (2D) simulation of a loosely coupled GNSS/INS, which is illustrated in Figure 5. 

GNSS position updates are used for INS drift mitigation. Acceleration due to gravity is in the direction opposite to the z-axis. The platform has a constant absolute velocity value of 20 m/s. The trajectory starts with a straight motion segment, follows it by a climb (with turning of the IMU body-frame), and completes with a straight motion. 

Inertial sensor errors were simulated as follows:

• Gyro drift: first-order Gauss-Markov process, standard deviation is 50 deg/hr, correlation time is 1000 s;

• Accelerometer bias: first-order Gauss-Markov process, standard deviation is 1 mg, correlation time is 1000 s.

GNSS position errors were simulated as zero-mean Gaussian processes with a standard deviation of 1 cm. This position performance corresponds to a carrier phase-based RTK solution. INS and GNSS update rates were 100 Hz and 1 Hz, respectively.

Figure 6 shows estimates of angular errors (attitude and gyro drift) and accelerometer biases. 

As shown in Figure 6, attitude error and x-accelerometer bias cannot be estimated separately (residual estimation errors remain) during the initial straight segment. Their estimates converge to true values after the climb starts and sufficient motion dynamics are accumulated for the error state separation. 

The plots provide initial insight into the influence of motion dynamics on the observability of INS error states. As shown in Figure 6, there is a residual attitude error that remains uncompensated during the straight motion phase. The estimate of the z-accelerometer bias rapidly converges to its true value while its x-accelerometer counterpart cannot be estimated. Both attitude error and x-accelerometer bias estimates converge to their true values shortly after the climb starts at about 30 seconds into the simulation.

This phenomenon can be explained as follows. Essentially, the Kalman filter numerically differentiates position error states into acceleration error and then separates the latter into attitude and bias error terms. For simplicity, consider a 2D case where (i) body-frame is aligned with the navigation frame; and, (ii) accelerometer biases, b, and angular orientation error, δα, are constant. INS does not take advantage of the known angular orientation (i.e., the fact that body and navigation frames are aligned with each other) and integrates gyro measurements into attitude. In this case, the navigation-frame acceleration error is:

For constant-velocity motion:

The z-bias component can be directly estimated from the z-acceleration error, which, in turn, is estimated from position error observations. However, observations of the x-acceleration error over time are rank-deficient. This does not allow for the separation between bias and attitude error terms. The filter balances their contribution into acceleration error but cannot estimate them individually. 

When time-varying acceleration is applied along the z-axis, the observation model in Equation 8 changes into:

When this system is observed over time, it has a full rank and bias and attitude error terms can now be estimated separately. This is what happens in Figure 6 after the platform starts climbing. 

Clearly, when centimeter accurate GNSS position is available all the time, it is not critical to estimate individual INS error terms. However, the observability aspect becomes more critical when GNSS outages are present. To illustrate, we consider two outage scenarios as shown in Figure 7. 

Outage 1 starts before the climb when attitude and bias errors cannot be separated. Outage 2 starts during the climb after the sufficient error state convergence is achieved. Figure 8 compares GNSS/INS position performance for these two outage cases. 

As shown in Figure 8, INS drift during the second outage is reduced significantly due to the ability to separate attitude and bias error states during the climb maneuver. Particularly, the maximum error growth is reduced (from 2.7 m and 5 m for x and z position components to 0.5 m and 2 cm, respectively) for the second outage scenario. This error reduction is due to the ability to separately estimate angular and linear INS error terms by using the motion dynamics.

Conclusion

Complementary sensor fusion is the most efficient approach to fuse INS with aiding sensors. The methodology (i) models inertial error dynamics rather than modeling full motion states, and (ii) applies differences between actual measurements and their INS-based predictions as estimation observables. One of the key benefits is the ability to use computationally efficient linear estimation techniques (such as an extended Kalman filter) for the error-state propagation. Another benefit is a straight-forward choice of the system noise matrix (that is fully defined by INS error models) without the need to make a trade-off between the system’s ability to suppress the measurement noise and accurately follow a platform’s motion (e.g., during a straight flight and a turn maneuver) dynamics. This column also illustrated that the observability of inertial error states is generally impacted by the motion dynamics, which can have substantial influence on navigation performance during GNSS outages. 

The post The Inertialist: Complementary Sensor Fusion appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

The prototype, consisting of a single Tx-Rx communication link, was tested in indoor-to-indoor, outdoor-to-outdoor and outdoor-to-indoor communication scenarios and in both static and dynamic conditions. The outdoor-to-indoor results indicate that, with further development, the Civilian and Assets Recovery System (CARS) is a promising navigation and positioning technology.

WAHYUDIN P. SYAM, DAVID SCOTT, ALEJANDRO PÉREZ CONESA, GMV

IGNACIO RODRÍGUEZ, MELISA LÓPEZ LECHUGA, ENRIC JUAN MARTINEZ, DEPARTMENT OF ELECTRONIC SYSTEMS, AALBORG UNIVERSITY

RIGAS THEMISTOKLIS IOANNIDES, EUROPEAN SPACE AGENCY

Crisis modes, following natural or human made disasters, can happen anywhere and any time, with response teams relying heavily on the limited telecommunication and localization infrastructures available to support rescue operations. The availability of navigation and communication capabilities in these situations is vital both for distressed users and core operating rescue teams, as obtained position accuracies will determine the effectiveness of the recovery. 

In these situations, a number of challenges must be overcome to provide resilient and reliable navigation. These include no access to power supplies, non-predictable environments, and the lack of operational terrestrial infrastructures. These challenges are problematic for GNSS-based or terrestrial-based positioning [1]. In addition, in difficult scenarios indoors or in urban canyons or catastrophes resulting from avalanches, floods or war, current navigation solutions, especially GNSS-based ones, are not adequate to support the required navigation functionalities.

Current Crisis Navigation Systems

There are several available communication and navigation systems available for crisis recovery and emergency scenarios. They include: 

Walkie-talkies: A half-duplex communication system that uses low frequency radio signals [2] and is the basic tool for communication during an emergency. 

Amateur radio: Radio communication that works within an amateur-allocated signal spectrum and is established via the ionosphere for wide range coverage that is useful in emergency situations [3]. 

Trunking radio: A direct communication system that relies on a computer-controlled network to manage channel assignments among different users. It has a spectral efficiency to accommodate voice and low-speed data, which is critical during a crisis. [4]. 

Cell on wheels: A vehicle-based mobile cell communication system where communication tower and trans receiver equipment are mounted. A deployable cell can connect to a main communication network via satellite backhaul [5], quickly establishing a 700 MHz frequency band with several kilometers of coverage for first responders. 

Mobile ad-hoc-network: A communication system based on ad-hoc Wi-Fi networks with limited communication range [6]. A smartphone base mobile ad-hoc system framework that leverages available Bluetooth/Wi-Fi/Sound modules to complement hardware-based mesh networks has been presented [7]. 

Base station ad-hoc-network: A communication system built from disconnected base stations that can connect to cellular networks [8]. 

Wireless mesh networks: These consist of a self-healing, self-configuring multi-hop wireless network with different types of architectures with limited coverage [9]. 

UAVs and balloon-based wireless systems [10]: Drones have been used as mobile nodes to extend the coverage of existing communication systems, established using the UHF and VHF frequency bands, to provide connectivity to users located in disaster areas [11,12]. A prototype system for a balloon network using two wireless nodes in the sky to establish emergency communications through an access point and a repeater has been proposed [13]. This balloon-based system has been tested on indoor and outdoor environments and shows a coverage range of up to 2,400 m. 

Device-to-device (D2D) cooperative communication: A system that allows communication between nearby mobile devices by establishing cellular or ad-hoc links [14]. The use of these systems has been demonstrated for emergency communication [15]. 

For outdoor navigation and positioning, GNSS has become common, with its applications widespread and accessible by most people with smartphones. GNSS is based on satellite constellations in space (commonly at medium-Earth orbit) that emit signal containing pseudorange, orbit and timing that are used to reconstruct the position of GNSS receivers on Earth [16]. GNSS has a global coverage, is in microwave frequency range and is effective for outdoor use, especially in open sky conditions. However, GNSS signals are weak and vulnerable to obstacles such as walls, roofs and tunnels. Examples of well-known global GNSS constellations are GPS [17], GLONASS [18], GALILEO [19] and BEIDOU [20].

There are indoor positioning and navigation systems based on radio frequency, optical, acoustic, magnetic signals and hybrid indoor positioning methods such as sensor fusion like LiDAR and inertial sensors. Examples of radio frequency indoor positioning systems include identification (RFID), Bluetooth low energy (BLE), Wi-Fi, Zigbee and ultra-wideband systems [21]. Optical signal-based positioning processes optical signals in infrared and visible light frequencies [22,23]. Photo-diode sensors can capture reflected lights that contain data, such as position information. Other less popular technologies leverage acoustic and magnetic signals for low-range indoor positionings [24,25]. 

Most of the aforementioned communication and navigation systems have drawbacks that make it difficult to use them in an emergency. For example, most only provide communication data and are not capable of providing navigation and localization (or otherwise), for either outdoor only or indoor only communication. They’re also large and heavy, aren’t portable, require an electrical power supply, require communication cell towers and other terrestrial infrastructures, have weak signal power and signal propagation capabilities (such as GNSS), have a small coverage area, need specialized equipment and trained users (such as amateur and trunking radios), can only be used indoors or outdoors, and are only for static use. In a natural disaster, a versatile, stand-alone and flexible communication system for positioning and navigation that can be used for outdoor-to-outdoor and outdoor-to-indoor communications in both static and dynamic conditions as well as in harsh environments is needed.

The first stage development of a flexible and fast-deployable Civilian and Assets Recovery System (CARS) that provides positioning capabilities in crisis situations and overcomes these challenges is presented in this article. The system accounts for these limitations, such as outdoor-to-outdoor and outdoor-to-indoor communication capability, and considers addressing further challenges, such as the need to cover wider areas, the need for lightweight devices at transmission and reception, and the need for flexibility and configurability to support operations in unpredictable hostile conditions. 

Figure 1 illustrates the importance of CARS in an emergency. In Figure 1, many victims may be trapped inside wrecked buildings or underground after a natural disaster. CARS must be flexible, deploy quickly and provide good signal propagation capabilities (outdoor-to-indoor) to transmit navigation signals to receivers that are underground or inside the wrecked buildings. From Figure 1, the CARS transmission (Tx) unit should be lightweight, battery-operated, easy to mount on many types of fast-deployable moving or flying vehicles, and able to send navigation data obtained from open-sky GNSS signals or internal inertial sensors. The receiver (Rx) unit should be portable and lightweight.

The first-stage CARS prototype consists of a single Tx-Rx communication link and is built on the high-configurability and flexibility of a software defined radio (SDR) device. This system emits signals that contain ranging information and has the propagation capability advantages that low frequency radio frequency (Low_RF) signals provide. The transmitter of the Low-RF system is based on a battery-powered SDR that can be mounted onto any terrestrial moving or flying vehicle, such as cars, trucks, cell-on-wheels and drones.

Radio Signal Propagation in a Cluttered Environment 

CARS must exploit good signal propagation capabilities to improve the system’s area coverage. Low frequency (sub-GHz) transmissions have lower outdoor propagation and penetration losses into buildings than higher frequency transmissions [26, 27], especially in a harsh environment [28]. This low propagation and penetration loss will increase the ability to transmit and receive signals in a cluttered environment, such as urban canyons and indoor/multi-floor transmission. In addition, with operation at narrow signal bandwidth, the low power requirement for the transmission will be obtained. An outdoor-to-indoor RF propagation in an urban-cluttered scenario is used to evaluate and study the RF propagation capability for the developed first-stage CARS protoype.

The outdoor-to-indoor RF signal propagation is presented in Figure 2. In this scenario, the Tx unit is always outdoors on a moving or flying vehicle. The Rx unit is in a building and obscured by walls and windows. The outdoor propagation loss can be modeled over the link from the Tx unit to the external facades of the building, and to the Rx unit behind walls inside the building. In Figure 2, d_out is the linear distance from the Tx unit to the incident point on the building façade as specific incident angle θ. d_in is the linear distance from the incident point on the facade to the Rx unit passing, in this case, two walls (n_walls=2). In this work, d_out is expected to be up to 2 km as the design choice of this case study. The RF signal propagation scenario shown in Figure 2 considers all conditions impacting the propagation loss: outdoor-to-outdoor, outdoor-to-indoor and indoor-to-indoor. Hence, this scenario can accuratley evaluate the proposed first-stage CARS for outdoor-to-indoor communication. In Figure 2, the urban clutter affects the true incident angle at the façade point and the height of the Tx unit affects the signal propagation trajectories. The higer the Tx unit height from the ground, the less effect the urban clutter will have on signal propagation loss.

The main propagation mechanisms considered are propagation in-between buildings [29], propagation above rooftops and multi-screen diffractions at the rooftop edges [30,31], external facade penetration loss through low-attenuating window and loss due to the grazing incident angle [32], and propagation loss due to guided indoor propagation and walls/doors/multi-floor openings [33]. The total propagation-loss (PL_out-to-in) of the outdoor-to-indoor transmission is formulated as:

Where PL_out is the propagation loss for outdoor-to-outdoor transmission and is a function of Tx height h_TX, Rx height h_RX, the linear distance from the Tx to an incident point of a building d_out and transmission frequency f. L_BEL is the propagation loss for outdoor-to-indoor transmission and is a function of building materials B_m, transmission frequency f and incident angle θ. PL_in is the propagation loss for indoor-to-indoor transmission and is a function of the linear distance from the incident point on the building to the Rx unit d_in and number of openings n_walls, such as walls, doors and corridors.

Different building materials will significantly affect the propagation loss of RF signals when entering buildings. Losses for several materials are estimated at different frequencies (100-500 MHz) from published data [35,36,37]. In this article, two types of buildings are considered: traditional (constructed mainly of red bricks) and non-traditional thermal efficient (constructed mainly of thermal-soaked glasses and steel beam structures). These building types have different signal propagation attenuation because of their constructing materials, making it important to evaluate the propagation characteristic of transmissions of the proposed system for both.

Low-RF Communication System Development 

SYSTEM ARCHITECTURE

The architecture of the Low-RF CARS consists of a transmitter/Tx (CREAM) unit and receiver/Rx (DREAM) unit that are based on reliable and highly configurable SDR devices. The system operates at a low frequency of 113 MHz to 500 MHz (UHF and VHF bands) and emits signals containing ranging and velocity information. The Tx unit is based on a compact SDR device that is battery-operated, can operate in stand-alone mode and can be mounted in various moving or flying vehicles, such as a UAV, cell-on-wheel, cars and trucks.

Figure 3a shows the Low-RF system architecture. The Tx unit operates using a battery and has an embedded CPU, FPGA, on-board GPS and IMU sensors. The CPU will prepare the message bit and pseudorandom (PRN) code and then modulate the signal. The prepared signal is then sent to the FPGA system for transmission. The Tx unit also can transmit a replayed signal from files. Although the Tx unit can operate in stand-alone mode, it requires a PC connection to start the operation of the Tx unit for the first time. The Rx unit is a portable SDR and needs to be connected to a host-PC to process received signals, although it could also be a laptop if device specifications are powerful enough for processing. An external clock can be connected to the Rx unit for synchronization purposes if required. All signal processing, including tracking, acquisition and demodulation, is carried out by the host-PC connected to the Rx unit. In general, the Low-RF system can use any suitable SDRs and signal designs.

Figure 3b shows the SDR uses in the Low-RF system. The Tx unit uses Ettus USRP E312 [40] and the Rx unit uses Ettus X310 [41] devices. The E312 features a Xilinx Zynq 7020 SoC (7 Series FPGA with ARM Cortex A9 866 MHz dual-core processor) and an Analog Devices AD9361 RFIC as the RF front end. This Tx unit supports an instantaneous bandwidth up to 56 MHz and frequency range between 70 MHz and 6 GHz. The X310 is equipped with a daughterboard UBX160 as a RF front-end and features a large Xilinx Kintex-7 FPGA (XC7K410T), sampling rate up to 120 MHz and frequency between 70 MHz and 6GHz, multiple high-speed interfaces including dual 10GbE, PCIe Express and dual 1 GbE. The flexible clocking architecture allows for a configurable sampling rate and synchronization with the optional high accuracy GPS disciplined oscillators (GPSDO). RF Network on Chip (RFNoC™) FPGA development framework and Ettus UHD are used for software development. Figure 4c shows two types of antennas used for wireless transmission: RETEVIS RT20 and RETEVIS RT1/3. The two antennas are used for transmission at 113 to 500 MHz. Table 1 presents their detailed specifications.

SIGNAL DESIGN AND MESSAGE BIT STRUCTURE

The implemented signal design considers the limited processing capability of the Tx unit due to the trade-off of having low-power consumption and battery-operated capabilities. The processing power is far below a standard workstation, so the unit’s signal design should be simple and efficient. The implemented signal should follow the structure of the spread-spectrum CDMA signal and implement a GPS L1 C/A-like signal structure, transmitted at low frequency 113 MHz to 500 MHz, with a baseband bandwidth of 1.023MHz, following the PRN chip rate as 1.023 Mcps [38]. Navigation data, containing position and velocity, uses binary-phased shifted keying (BPSK) modulated in In-phase and Quadrature (I/Q).

The navigation or message data contains 200 bits that are 8-bit preamble, 16-bit message id, 34-bit for each Earth-center-Earth-fixed (ECEF) X, Y, Z position, 16-bit for each ECEF X, Y, Z velocity and 16-bit cyclic-redundancy check (CRC) code as shown in Figure 4. The selected number of navigation bits is a trade-off between numbers of information to contain and the unit’s hardware limitation. These bit structures use the ECEF coordinate reference system for easy translation to any other coordinate system, have a high-resolution positioning and use message ID for message tracking and verification and Tx-Rx synchronization. The position granularity (resolution) of the navigation bit is computed as follows: 

Because the position (in meters) is 24 bits to represent the integer part of the position and 10 bits to represent the decimal point of the position, the smallest value the position bit can represent is when all the bits are zero, except the least significant bit (the 34th bit), is 1. Hence, when only the least significant bit is 1, the bit will represent 1/210, that is the position granularity is approximately 0.000976 m = 0.976 mm. 

NAVIGATION MESSAGE PREPARATION AND MODULATION

It is important to note that unlike receiving, where the sampling frequency should not be a multiple integer of the data rate (at a specific input frequency), the sampling frequency for transmitting should be an integer multiple of the data rate [39]. For transmitting the navigation signals, it’s necessary to up-sample the 1.023 Mcps signals to be at least 2 Mcps and the integer multiple of the sampling rate of transmissions. The signal should be up-sampled at least to 2 Mcps before being sent to digital-to-analog (DAC) processing of the Tx unit. To up-sample the signal, the PRN code is processed per one period of 1 ms. The PRN used has 1,023 samples per 1 ms. From these 1,023 samples, an up-sampling process is applied. The up-sampling process increases the number of samples or chips of the one-period PRN from 1,023 samples per 1 ms (equal to 1.023 million samples per 1 s, that is non-integer number per million) to 2,000 chips per 1 ms. By doing this up-sampling, the number of samples of the PRN becomes 2 million samples per 1 s period (integer number per million). Because the up-sampled PRN chips are not exactly 2× (that is 2,000 samples per 1 ms) of the base PRN chips (1,023 samples per 1 ms), there are chips that are multiplied by either 1× or 2×.

For the up-sampled PRN with 2,000 samples per 1 ms, two sets of bit strings are prepared, one with modulo sum 2 with 0 bit (for 0 bit navigation data) and another with modulo sum 2 with 1 bit (for 1 bit navigation data). Because the navigation data has 200 bps, the period of each bit is 5 ms. Hence, the whole navigation message is prepared for every 5 ms. If the navigation data bit is 0, then the set of bit strings that has been modulo summed 2 with 0 bit are used. Finally, all samples of navigation data for 1 s period (2 million samples) are BPSK modulated by changing the 0 bit value into 1 and 1 bit value into -1. From the 2 Mcps of the up-sample signals, these signals are then sent to the TX unit’s DAC at a sampling rate of >2×2M = 4 Mcps. In our case, the Tx sampling rate is set to be 6 Mcps (MHz). This 6 Mcps sampling rate is chosen by experiments. The experiments perform transmission-receiving tests with various different up-sampling rates and transmission sampling rates to find the maximum values the Tx unit can process.

RECEIVER SENSITIVITY

This is estimated experimentally in laboratory conditions. The sensitivity test is performed by reducing the signal power emitted by the Tx unit until the signal can no longer be acquired by the Rx unit. The transmission between the Tx and Rx unit uses a cable connection to avoid any external disturbance. When received signals cannot be acquired, the signals cannot be tracked and subsequently cannot be demodulated. The sensitivity analysis starts by transmitting Tx unit signals at minimum power, which in this case is -43.7 dBm. This power level accounts for the signal generated by the Tx unit plus any attenuation caused by the cable connection. Attenuation is progressively added at the receiver’s input until the signal processing can’t acquire the signals received by the Rx unit. Based on these tests, the Low-RF system sensitivity is between -115 dBm and -120 dBm at 10 MHz bandwidth. It is important to note these sensitivity values are the best possible sensitivity, as attenuation from channel impairments is minimal because of the cable connection. The sensitivity is expected to be lower than the sensitivity obtained from the lab test in real-world scenarios, as more factors such as channel impairments, phase coupling and other signal propagation losses add to the attenuation.

Low-RF Wireless Transmission Experiments in Different Scenarios 

This section presents the results of wireless transmission tests performed both in laboratory, indoor, outdoor (system initial test) and real-world scenarios. To be classified as a successful transmission, the demodulated data at the Rx unit must be correct or match the data sent by the Tx unit. This main criterion is quantified as the success rate, that is the number of correct demodulated data at the Rx unit per total number of transmitted data by the Tx unit. The X,Y,Z positions and velocities in the ECEF coordinate system are obtained from the GPS sensor in the Tx unit and then fixed for the transmission tests. In all experiments, the sampling rate at the Rx unit is 10 MHz. Before real-world scenario tests, system initial tests are performed to verify the end-to-end signal processing chain and to select the best frequency for real-world scenario experiments.

SYSTEM INITIAL TEST EXPERIMENTS:

Controlled transmission with physical channel and closed wireless transmission

This test validates the signal modulation and demodulation and the antenna characteristics. The activities test the whole end-to-end signal processing chain from navigation message preparation and modulation until the signal receiving, acquisition, tracking and demodulation processes. Figures 5a and Figure 5b show the wired-transmission setup and the wireless transmission at a close line-of-sight of 0.3 m distance. The signal received from the wired transmission at 400 MHz contains minimal noise. The baseband signals have higher power than noise as well as a strong correlation peak (Figure 7-top). The signal tracking processes are very stable (Figure 8-top). Note that, although both the Tx and Rx are static, there is a Doppler frequency up to 400 Hz due to TCXO clock noise, that is commonly 1-2 ppm, at the Rx unit [38]. Similar to the wired transmission, the close-distance wireless transmission (at 113, 144, 400 and 500 MHz) has low noise, except for some fluctuations during signal tracking processes. A success rate of 100% is obtained from the wired and close-distance wireless transmissions (Table 2).

Static indoor-to-indoor transmission

Figure 6a and 6b present indoor-corridor experiments performed at 8 m (corridor width = 2 m) and indoor-between-rooms at 10 m, between the Tx-Rx respectively. These experiments include channel impairments such as multipath and blockade effects. The transmissions consider different setup combinations for antenna 1 and 2 at 113, 133, 144, 400 and 500 MHz. Results from these experiments show a 100% transmission success rate as presented in Table 2. In some cases, the channel impairments add significant noise to the transmitted signals, especially in the indoor-between-room transmissions, where the baseband signals are buried under the noise (Figure 7-middle). Also, for tracking processes, more fluctuations on the carrier Doppler tracking are observed for indoor-between-rooms as multipath and reflection effects are stronger than the indoor-corridor scenario (Figure 8-middle). In these tests, the Doppler is about 500 to 750 Hz.

Static outdoor-to-outdoor transmission

Figure 5c shows the outdoor-free path experiments with both antennas at 113, 133, 401.5 and 500 MHz. The distance between the Tx-Rx is set to be approximately 60 m. Transmission at 113 MHz only shows a 33% success rate so it is concluded that this frequency is not effective for real-word testing. This is because 113 MHz is well outside the main radiation band of the setup antennas, which limits the efficiency of the system at this frequency. Other frequencies show a 100% success rate, except 401.5 MHz, which shows a 85% success rate. From Figure 7-bottom, it can be observed that this free-path transmission has low noise power compared to the baseband signal power (the rectangular waveforms are still observable). There are small fluctuations of the carrier tracking, shown in Figure 8-bottom, which may be caused by power loss variation.

All the correlation peaks of the signal acquisition, for 1 ms correlation length, show obvious triangle shapes indicating the signal correlation can effectively acquire the PRN code from the received signal. The selected 1 ms of correlation length is to speed up computation process and avoid data bit transition during correlations. We decided to bring 133 MHz (with antenna 1), 401.5 MHz (with antenna 2), and 500 MHz (with antenna 2) to the real-world experiments. 

Real-world scenario experiments

Low-RF real-world experimentations were performed at the Harwell Science and Innovation Campus (UK), which is a suburban area with a low density of two to three story buildings, single-lane roads, and is densely vegetated. Two types of buildings were considered for outdoor-to-indoor transmission: traditional (constructed of red bricks and 40 cm concrete walls) and thermal efficient (constructed of thermally soaked glasses and a steel-beam with 36 cm thickness). Table 3 shows the setup parameters of the experiments for the Tx and Rx units, which are transmitted signal power, antenna gain and effective receiver sensitivity.

Before the real-word experiments, calibrations were performed to estimate the effective signal penetration loss for the two buildings. A series of received power measurements by the Rx unit at different indoor positions were performed for calibrations. The Tx unit is fixed at a 1.9 m height and a distance of 50 m and 13 m from the traditional and thermal-efficient building, respectively (Figure 9). For the traditional building, transmissions are at 133, 401.5 and 500 MHz and for the thermal-efficient building 133 and 500 Mhz. A total of 43 tests were performed considering the different buildings, Rx positions inside the buildings and transmission frequencies, so a reliable estimation of penetration losses could be obtained. 

The results of the building penetration loss reference measurements correspond to 10.4 dBm (at 133 MHz), 16 dBm (at 401.5 MHz), and 8 dBm (at 500 MHz) for the traditional building; and 15.2 dBm (at 133 MHz), and 25.2 dBm (at 500 MHz) for the thermal-efficient building. The penetration loss experienced in the thermal-efficient building is higher than the traditional building for about 50%-200%. The penetration loss for the traditional building does not follow the expected trend (the higher the frequency, the higher the loss). This is because there is a high statistical variability component for penetration loss in low-attenuation buildings (i.e. traditional) because of materials used and structure. For thermal-efficient buildings, the penetration loss values match the expected radiofrequency behavior.

Static outdoor-to-indoor transmission: Vertical sweep

This test evaluates and validates the operational range of the Low-RF system by considering multiple Tx height topologies emulating different CARS platforms (2.5 m: emulating a TX mounted on a car; 7.5 m: emulating a lamppost-mounted Tx; 20 m: emulating a Tx deployed on top of a building; 75-120 m: emulating an UAV-mounted Tx) at distant locations from the two buildings. The distance between the Tx-Rx unit is from 500 m to 2 km as shown in Figure 10a. The Rx unit is placed at outdoor and indoor positions. Sixty tests were performed considering the two buildings, Tx-Rx positions and transmission frequencies.

The highest and lowest received signal powers for the traditional building at the indoor location and 2 km Tx-Rx distance are -59 dBm and -61 dBm, -81 dBm and -98 dBm, -85.9 dBm and -94 dBm at 133, 401.5 and 500 MHz, respectively. For the thermal-efficient building, the highest received signal powers at 590 m Tx-Rx distance are -50 dBm -70 dBm for 133 and 500 MHz, respectively. The highest received powers are obtained when the Tx (mounted on the drone) is at 120 m. As expected, the received powers at indoor locations are lower than outdoor locations. A better signal power is generally observed at higher heights due to the decreased presence of low-height obstacles (vegetation, foliage, buildings). Figure 11 shows visual comparisons of signal noise during acquisition, tracking and demodulation when the signal is transmitted with 500 MHz at 20 m and 100 m height where the Tx-Rx distance is 2 km. Figure 11 also illustrates that transmissions at 100 m experience lower signal noise than those at low height.

The success rates of the transmission for the traditional building for each receiver block processing at different Tx-Rx positions and frequencies are shown in Figure 12. For 133 MHz, the transmissions are nearly unsuccessful due to antenna and propagation limitations. For 401.5 and 500 MHz, the transmissions were mostly successful, especially at a high Tx (CREAM) position because there are fewer blockades. The transmissions at 500 MHz have the best performance. From Figure 12, it is clear transmissions at 12.5 m height have significantly higher failure rates compared to transmissions at > 20 m height due to obstacles imposing high signal attenuations up to below the Low-RF system sensitivity. It is worth noting that demodulation processes are also affected by near-far effect caused by multiple transmitters at different distances.

Results of the transmission tests for the thermal efficient buildings are shown in Figure 13. Like the traditional buildings, the transmissions at 133 MHz are not effective. However, transmissions at 500 MHz have a high success rate. For transmissions where the Tx is at low height and the Rx is outdoor, the transmission success rate is significantly lower than the results at > 20 m, due to signal attenuations introduced by the surroundings and the building materials. For transmissions where the Rx unit is indoor, only transmissions from the Tx unit at 120 m height maintain the success rate. Note the demodulation success rate of the transmissions at 120 m for both outdoor and indoor is low. However, because the acquisition and tracking processes are mostly successful, the failure of demodulation is caused by bit flipping. This could be corrected by implementing error correction techniques to improve the demodulation success rate. We found the transmitter CPU computational capacity limits the implementation of such techniques.

Dynamic outdoor-to-indoor transmission: LEO-pass trajectory

This test analyzes the Low-RF system performance under LEO satellite trajectories. The LEO satellite orbit pass is emulated by flying a UAV, carrying the Tx unit, at 40 m height in a straight line passing over the two buildings and by covering elevation angle α = [10°, 170°] as shown in Figure 10b. Thirteen positions were used to emulate the pass of a LEO satellite over the two buildings. For the traditional building, the Rx unit was located in four different positions: outdoor, indoor—first floor, indoor—first floor, and indoor—second floor. For the thermal-efficient building, the Rx unit is located at two locations: outdoor and indoor positions.

The transmission results for the emulated LEO orbit experiments passing the traditional and thermal-efficient buildings are shown in Figure 14a and 14b. In Figure 14a, the Low-RF system performance for the traditional building has the best performance at 500 MHz, with an acquisition and a tracking success rate of nearly 100%. The demodulation success rate was about 80% for all four positions. At 401.5 MHz, the demodulation success rate is slightly lower than others. That may be caused by interference with other signals at the frequency. The worst performances were at 133 MHz (especially at the indoor ground floor). This is not the design frequency of the antenna set, so signals are transmitted with less power. The thermal-efficient building had a success rate of nearly 100% for acquisition and tracking and about 30 to 50% for demodulation at 500 MHz. Regarding demodulation, due to less received power (large noise) at the Rx unit in the thermal-efficient building, the bit flipping percentage is higher than in traditional buildings. Again, one way to improve the demodulation process is to implement error-correcting techniques (channel coding) in Tx-Rx signals.

When it comes to received powers for experiments at the traditional building, the highest power is at outdoor locations and the lowest at the indoor ground floor, as expected. The highest and lowest received powers are -20 dBm and -55 dBm, -35 dBm and -75 dBm, and -45 dBm and -85 dBm, and for frequency 133, 401.5 and 500 MHz, respectively. The highest received powers are mostly at elevation angles between 10° and 60° where transmitted signals can reach the Rx through windows, thus without suffering wall attenuation. For the thermal efficient building, the highest (outdoor) and lowest (indoor) received powers correspond to -22 dBm and -55 dBm, -30 dBm and -75 dBm, and for frequency 133 and 500 MHz, respectively. Similar to the traditional building, the highest received powers are mostly at elevation angles between 10° to 60°.

From the received Rx power from the LEO pass experiment emulated by the flying drone at 40 m, the power loss from LEO satellites can be extrapolated by an extra path loss factor to account for the expected propagation from a LEO satellite flying at 500 km. By taking into account the realistic elevation geometry and distance (with respect to the emulated LEO-pass test), the extrapolation of the Rx power from LEO orbit is calculated as: received Rx power (at 40 m, elevation angle) – 20×log₁₀ (distance between the Tx to LEO, elevation angle). From the extrapolation, all estimated received LEO RF power levels would be above the sensitivity of an ideal GPS L1 C/A receiver with processing gain PG = 25 dB (-158.5 dBm). All estimated values are based on the Tx-Rx parameters presented in Table 3.

Conclusion and Future Work

The Low-RF system, based on a single Tx-Rx communication link, has successfully demonstrated possible outdoor-to-indoor transmissions both in static and dynamic modes. Promising results were obtained from transmissions at 133 MHz, 401.5 MHz and 500 MHz. With the combination of explored hardware and considered experimental scenarios, the best signal propagation and navigation capabilities are obtained with the 500 MHz transmission. From the demonstration, the developed system, with navigation signals at low frequency (VHF-UHF frequency), efficient spread-spectrum signal structure and flexible Tx-Rx architectures can be used for a full CARS implementation and offers huge potentials for users such as emergency service responders. From real-world scenario experiments (vertical sweep and LEO pass), results show using low frequency signals for CARS is promising because of their improved penetration capability and ability to carry navigation information, which is critical in emergencies. From the results of the LEO pass experiments, it is possible to extract the feasibility of LEO-PNT systems making use of low-frequency transmissions.

Future works include extending transmission testing to other frequencies (C, S and high UHF band), testing in more scenarios (e.g. different types of buildings, urban cluttered areas, vegetations, Tx-Rx topologies and dynamic tests where the Rx moves), testing with different signal design and modulation, for example using a shorter PRN to reduce the spectrum bandwidth, and improving the Low-RF platform by using at least four Tx units, reducing the size, weight and power of the system and increasing the technology-readiness level (TRL) for deployment within an operational environment. 

Acknowledgments

This work was supported by the European Space Agency (ESA) NAVISP program Element 1, which aims to innovate the technology of PNT systems. The work described was done under an ESA Contract. Responsibility for the contents resides in the author or organization that prepared it.

We would like to thank David Payne (GMV) who was the drone operator for the duration of the experimentation campaign, Aron Kisdi (GMV) who supported the drone experiments, Tim Whitworth and Yeqiu Ying who were involved in the technical activity at the early phase of this project and Filipa Fernandes for her contributions to reviewing the state of the art. Finally, we would also like to thank Mark Dumville (GMV) for his continuous support throughout the activity.

This article is based on material presented in a technical paper at ION GNSS+ 2022, available at ion.org/publications/order-publications.cfm.

References

(1) Huang, J.S. and Lien, Y.N., 2012, November. “Challenges of emergency communication network for disaster response”, IEEE International Conference on Communication Systems (ICCS), November 2012, pp. 528-532.

(2) Srinivasa, D., Pradep, P. D., Kumar, B. A., “Survey of Emergency Communication Network Architectures,” International Journal of u-and e-Service, Science and Technology, vol. 8, 2015, pp. 61-68.

(3) Nollet, K.E. and Ohto, H., “When all else fails: 21st century Amateur Radio as an emergency communications medium” Transfusion and apheresis science, 49(3), 2013, pp.422-427.

(4) Kunavut, K., “An Overview of Digital Trunked Radio: Technologies and Standards.” Journal of Industrial Technology, 10 (2), 2014, pp.111-121.

(5) FirstNet Authority report, “Understanding the FirstNet Deployable Program High-impact solutions for public safety operations,” 2021, USA.

(6) Mounir, T.A., Samia, M., Mohamed, S. and Bouabdellah, K., “A routing Ad Hoc network for disaster scenarios.” 1st International Conference on Information and Communication Technologies for Disaster Management, pp. 1-8, 2014, March.

(7) Rengaraju, P., Sethuramalingam, K. and Lung, C.H., “Providing Internet Access for Post-Disaster Communications using Balloon Networks.” Proceedings of the 18th ACM Symposium on Performance Evaluation of Wireless Ad Hoc, Sensor, & Ubiquitous Networks, pp. 111-117, 2021, November.

(8) Kishorbhai, V.Y. and Vasantbhai, N.N., “AON: a survey on emergency communication systems during a catastrophic disaster.” Procedia computer science, 115, 2017, pp.838-845.

(9) Portmann, M. and Pirzada, A.A., “Wireless mesh networks for public safety and crisis management applications.” IEEE Internet computing, 12(1), 2008, pp.18-25.

(10) Huang, J.S. and Lien, Y.N., “Challenges of emergency communication network for disaster response.” IEEE International Conference on Communication Systems (ICCS), 2012, November, pp. 528-532.

(11) Hussain, S.A., Ahmad, N., Latiff, L.A., Ahmed, S.B.M. and Mohamed, N., “Disaster Area Network Expansion Using Drones Based Ad-hoc Cellular Communications.” IEEE Symposium On Future Telecommunication Technologies (SOFTT), pp. 22-27, 2021, December.

(12) Shamsoshoara, A., Afghah, F., Blasch, E., Ashdown, J. and Bennis, M., “Uav-assisted communication in remote disaster areas using imitation learning.” IEEE Open Journal of the Communications Society, 2, 2021, pp.738-753.

(13) Wang, Y., Su, Z., Zhang, N. and Fang, D., “Disaster Relief Wireless Networks: Challenges and Solutions.” IEEE Wireless Communications, 28, 2021, pp.148-155.

(14) Fodor, G., Parkvall, S., Sorrentino, S., Wallentin, P., Lu, Q. and Brahmi, N., “Device-to-device communications for national security and public safety.” IEEE Access, 2, 2014, pp.1510-1520.

(15) Chu, Z., Le, T.A., Nguyen, H.X., Nallanathan, A. and Karamanoglu, M., “A stackelberg-game approach for disaster-recovery communications utilizing cooperative D2D.” IEEE Access, 6, 2017, pp.10733-10742.

(16) European Global Navigation Satellite Systems Agency, “What is GNSS?” link: https://www.euspa.europa.eu/european-space/eu-space-programme/what-gnss#:~:text=Global%20Navigation%20Satellite%20System%20(GNSS,definition%2C%20GNSS%20provides%20global%20coverage.

(17) U.S. Department of Defense, “GPS Standard Positioning Service (SPS) Performance Standard,” GPS NAVSTAR, 2008.

(18) European Space Agency, “GLONASS Space Segment,” ESA Navipedia, 2018.

(19) European Global Navigation Satellite Systems Agency, “Galileo: the European global satellite-based navigation system,” European GSA Online, 2018.

(20) China Satellite Navigation Office, “BeiDou System Overview,” 2018

(21) Brena, R.F., García-Vázquez, J.P., Galván-Tejada, C.E., Muñoz-Rodriguez, D., Vargas-Rosales, C. and Fangmeyer, J., “Evolution of indoor positioning technologies: A survey.” Journal of Sensors, 2017.

(22) Karunatilaka, D., Zafar, F., Kalavally, V. and Parthiban, R., “LED based indoor visible light communications: State of the art.” IEEE communications surveys & tutorials, 17(3), 2015, pp.1649-1678.

(23) Haruyama, S., “Visible light communications.” In 36th European Conference and Exhibition on Optical Communication, 2010, September, pp. 1-22.

(24) Ijaz, F., Yang, H.K., Ahmad, A.W. and Lee, C., “Indoor positioning: A review of indoor ultrasonic positioning systems.” 15th International Conference on Advanced Communications Technology (ICACT), 2013, pp. 1146-1150).

(25) Li, K., Gong, Q., Ren, Y., Li, Y., Han, Y., Pang, C. and Kong, H., “Magnetic Field Positioning Technology of Indoor Sports Bodies.” IEEE Sensors Journal, 22(1), 2021, pp.219-228.

(26) Holloway, C.L., Koepke, G., Camell, D., Remley, K.A., Williams, D.F., Schima, S.A., Canales, S. and Tamura, D.T., “Propagation and detection of radio signals before, during, and after the implosion of a 13-story apartment building,” National Institute of Standard Technology Note, 2005, 1540.

(27) Wassie, D.A., Rodriguez, I., Berardinelli, G., Tavares, F.M., Sørensen, T.B., Hansen, T.L. and Mogensen, P., “An agile multi-node multi-antenna wireless channel sounding system,” IEEE Access, 7, 2019, pp.17503-17516.

(28) He, J., Zhou, S.Q., Liao, K., Wei, X.C. and Li, E.P., “Radio frequency propagation characteristics in disaster scenarios,” In Proceedings of 2014 3rd Asia-Pacific Conference on Antennas and Propagation, 2014, July, pp. 818-821.

(29) 3GPP, “Study on channel model for frequencies from 0.5 to 100 GHz (Release 15),” 3GPP TR 38.901, 2018.

(30) European Cooperation in Science and Technology, “Evolution of Land Mobile Radio,” COST 231, Final Report, 1999.

(31) 3GPP, “Study on New Radio (NR) to support non terrestrial networks (Release 15),” 3GPP TR 38.811, 2018.

(32) Rodriguez, I., Nguyen, H.C., Sørensen, T.B., Zhao, Z., Guan, H. and Mogensen, P., “A novel geometrical height gain model for line-of-sight urban micro cells below 6 GHz,” In 2016 International Symposium on Wireless Communication Systems, 2016, September,pp. 393-398.

(33) Colpaert, A., Vinogradov, E. and Pollin, S., “Aerial coverage analysis of cellular systems at LTE and mmWave frequencies using 3D city models,” Sensors, 18(12), 2018, p.4311.

(34) Rodriguez, I., Nguyen, H.C., Kovács, I.Z., Sørensen, T.B. and Mogensen, P., “An empirical outdoor-to-indoor path loss model from below 6 GHz to cm-wave frequency bands.” IEEE antennas and wireless propagation letters, 16,2016, pp.1329-1332.

(35) International Telecommunications Union, “Effects of building materials and structures on radiowave propagation above about 100 MHz,” ITU-R Rec. P.2040-1, 2015.

(36) Koppel, T., Shishkin, A., Haldre, H., Toropovs, N., Vilcane, I. and Tint, P., “Reflection and transmission properties of common construction materials at 2.4 GHz frequency,” Energy Procedia, 113, 2017, pp.158-165.

(37) William, C.S., Gary, R.B. and Robert, E.H., “Electromagnetic signal attenuation in construction materials” USA: National Institute of Standards and Technology, 1997.

(38) Misra, P, Enge, P., “Global Positioning System: Signals, Measurement and Performance,” 2nd Edition, 2006, Ganga-Jamuna Press: USA.

(39) Gallager, R.G., “Principles of digital communication,” 2008, Cambridge: Cambridge University Press.

(40) https://www.ettus.com/all-products/usrp-e312/. Accessed: 20 June 2022.

(41) https://www.ettus.com/all-products/x310-kit/. Accessed: 20 June 2022.

Authors

Wahyudin P. Syam holds a Ph.D. in geometrical measurement and uncertainty analysis from Politecnico di Milano in Milan, Italy. After finishing his Ph.D., he held a post-doctoral research position at the University of Nottingham for five years, where he gained experience in machine learning for measurement and image analysis and hardware-interfacing programming. He joined GMV in February 2021. 

David Scott received his MEng in Electrical and Electronic Engineering from Newcastle University in 2019. He has worked at GMV since graduation, where he is involved in several projects related to low frequency communications and alternative PNT solutions. He holds extensive experience in the manipulation of software defined radio (SDR) devices, including universal hardware driver (UHD), libIIO and GNURadio. He also has experience with embedded microcontroller programming in multiple architectures.

Alejandro Pérez Conesa received his B.Sc. degree in computer engineering and his B.Sc. and M.Sc. degrees in telecommunication engineering from the Universitat Autònoma de Barcelona (UAB) in 2018, 2019 and 2020, respectively. He is Project Manager for GMV and an Adjunct Professor at UAB. He worked as research engineer for the Signal Processing for Communications and Navigation (SPCOMNAV) group at UAB in 2019 and 2020 and as GNSS Technical Lead for Indra in 2020 and 2021. He has been involved in numerous activities funded by the European Space Agency (ESA) and other international entities.

Ignacio Rodríguez received B.Sc. and M.Sc. degrees in telecommunication engineering from University of Oviedo, Spain, and a M.Sc. degree in mobile communications and the Ph.D. degree in wireless communications from Aalborg University, Denmark. He is a Ramón y Cajal Research Fellow at University of Oviedo, Spain. Previously, he was an Assistant Professor at Aalborg University, Denmark, where he led the 5G for Industries Research Group; and an External Research Engineer with Nokia Bell Labs, where he was involved in 3GPP and ITU-R standardization activities. He was a co-recipient of the IEEE VTS 2017 Neal Shepherd Memorial Best Propagation Paper Award, and in 2019, was co-awarded the 5G-prize by the Danish Energy Agency and the Danish Society of Telecommunication Engineers.

Melisa López Lechuga received a B.Sc. and M.Sc. in telecommunications engineering from Universitat Politécnica de Catalunya, Spain, in 2016 and 2018, respectively. In 2022, she received a Ph.D. in wireless communications from Aalborg University, Denmark, where she is currently a postdoctoral researcher. 

Enric Juan Martinez received a B.Sc. and M.Sc. in telecommunications engineering from Universitat Politécnica de Catalunya, Spain, in 2015 and 2018, respectively. He is currently pursuing a Ph.D. in wireless communications at Aalborg University, Denmark, in collaboration with Nokia, Aalborg, Denmark.

Rigas Themistoklis Ioannides received a B.Eng degree in electronic engineering from the University of Lancaster, Lancaster, UK, in 1995, a M.Sc degree on “Communication and Real Time Electronic Systems” from the University of Bradford, Yorkshire, UK, in 1997, and a Ph.D. from the University of Leeds, Leeds, UK, on modelling and correcting of ionospheric effects on GPS signals in 2002. He joined the European Space Agency in 2010. From 2014, he supported the Galileo Program on the definition and design of the Public Regulated Service. He joined the Galileo Second Generation in 2018, leading the design on OS Authentication techniques. In 2021 he joined the GNSS system Evolutions section under the Horizon Europe Program.

The post Navigating Emergencies with a Low-RF CARS appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.