“NAVGUARD is a compliment to GEODE,” said Frank Wilms, Principal at FDC and project coordinator for both the GEODE and NAVGUARD projects. “Where GEODE is developing mostly receivers for mobile platforms, the objective of NAVGUARD is to monitor threats to PRS signal integrity, to provide technologies to detect illegitimate activities in GNSS frequency bands and geolocate the sources of malicious activities. This will employ a range of system elements, which will ultimately provide a robust and trustable PRS-based PNT capability for our end user equipment.”

Full R&D program

Among the core elements being developed under NAVGUARD is a complete surveillance system, consisting of ground-based sensors and a space-based surveillance subsystem. “We will be prototyping a range of space payloads, one of which will be launched in the frame of the next four years,” Wilms said, “and there will be an information management subsystem which is at the center of the information platform. This will gather data on threats to PRS signal integrity and redistribute it to the users.”

User equipment, including new PRS mobile receivers, also to be prototyped under the NAVGUARD project, will form elements contributing to the collection of threat identification data. “The NAVGUARD endeavor will also deliver a number of other innovative technology elements whose purpose will be to increase PRS PNT robustness,” Wilms said.

NAVGUARD, just launched in February 2023, brings together 31 companies from 11 countries and will run for four years. Like GEODE, it encompasses operational system demonstrations, including the above-mentioned space payload in orbit. It is co-funded under the European Defence Fund (EDF) 2021 framework, comprising an industrial consortium that includes coordinator FDC, Thales, Leonardo, OHB, GMV, Aerospacelab and many others.

Speaking at the recent Munich Satellite Summit, Wilms said, “This is the first time the European Commission has granted us permission to talk about these projects in public. GEODE and NAVGUARD are the two biggest PRS defence projects, developing a common infrastructure for EU NAVWAR capability, relying on GNSS spectrum surveillance and extending PRS equipment integration platforms to bring us one more step towards a scaled and cost-effective procurement for our PRS end users.”

The post NAVGUARD for Ground – A Space-Based PRS Integrity appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

Speaking at the recent Satellite Summit in Munich, Frank Wilms, Principal at FDC, said, “The development of the Galileo Public Regulated Service (PRS) user segment is under the authority and is the responsibility of the Galileo member states. To that end, two large projects, GEODE and NAVGARD, have been launched.” The first of these, GEODE, was launched in 2021 and brings together 30 companies from 14 countries, including the likes of Airbus Defence and Space, Siemens, Orolia, Leonardo, Thales, GMV, Telespazio, Saab and many others. The goal of the project is to provide security modules, receivers and infrastructure prototyping.

“GEODE will feature operational field tests in various domains,” said Wilms, “for army, navy, drones and timing and synchronization applications. The sheer size of this project is indicated by the number, 1500, of deliverables and outputs. Our initial intention is to derive common technical specifications and standards, and use them to implement a range of equipment assets, to deliver fortified and operational PRS receivers, expected to be ready for demonstration in 2026.”

To make this happen, GEODE is developing a range of security modules, based on different technologies such as field programmable gate arrays (FPGAs) and application specific integrated circuits (ASICs). FPGAs allow customers the ability to reconfigure hardware to meet specific use case requirements after the manufacturing process, while ASICs are more power efficient and have better performance than off-the-shelf general purpose integrated circuits.

“These new GEODE security modules are mostly hidden behind the security umbrellas of the member states who are responsible for their development and their accreditation,” Wilms said. “Receiver and infrastructure prototyping will be taking place for standalone PRS receivers as well as server-based PRS receivers, for maritime PRS, timing and synchronization PRS, space server-based PRS and others, plus a certain range of support facilities. This includes a range of defense fields, all integrating these prototypes, to deployed on various land, maritime, air, RPAS [remotely piloted aircraft systems] and timing and sync platforms, for the end user demonstrations at the end of the project.”

Wilms pointed to a number of smaller member states participating in the GEODE project demonstrations, such as the Czech Republic, Greece and Romania. “From the user/procurement agent perspective, this is very important,” he said. “We want to to deliver to the member states equipment that is interoperable, ideally plug and play compatibility so that all member states have a choice of equipment, ready for use, that they can select on a case-by-case basis, for their own particular platform needs.” Among other things, the European Commission hopes to see GEODE boosting EU industrial competitiveness in the highly strategic domain of military positioning, timing and synchronization, while advancing the equipping of EU member states’ military forces with Galileo PRS capability.

The post GEODE PRS Project Biggest Ever Launched appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

One of the specific objectives of the new scheme is to improve incentives for transitioning the country’s HGV fleet towards lower emission vehicles. It will also help reduce the impact these vehicles have on infrastructure and road wear costs, as well as their noise impact.

The project is being managed by Sund & Bælt, a Danish government-owned company that manages the Storebælt and Øresund road links. The company will be responsible for both implementation and operation of the new system. A GNSS onboard unit (OBU), costing around €150 including installation, will be mandatory for HGV operators, with a cheaper, self-installed OBU option also planned. Manual or pre-paid ticket options will likely not be available.

No standing still

A first step in the procurement process has already been initiated and will continue until early 2024. Sund & Bælt have engaged in talks with a number of suppliers with experience in the delivery and maintenance of GNSS and distance-based tolling solutions. Key technical challenges include GNSS data handling, map and toll context data and map-matching, segment identification and toll calculation. A testing phase is taking place in early 2023, with final commissioning set for early 2025. Under the current Eurovignette scheme, HGV operators have to buy a small electronic device if they want to use motorways and toll highways in the Eurovignette countries, which include Denmark, Luxemburg, the Netherlands and Sweden. Eurovignette’s revenues are currently around €67 million per year. The Danish government expects the new GNSS-based scheme to match that until 2027, and then to double it from 2028 onwards.

The new scheme leverages significant technological advancements as well as increasing market maturity of GNSS telematics OBUs, mobile communications and enforcement equipment, all occurring during the past decade. The costs of establishing and operating the system will be comparable to those of previous non-GNSS-based schemes.

The post Denmark Planning New GNSS-Based Road User Charging Scheme appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

Signal design is one of the core tasks of GNSS development, because the broadcast signal is the only interface of the system for receivers, and its inherent performance determines the success of the entire system’s performance. Once the signal detail is defined, any subsequent change may cause changes to a large number of user terminals, which could have significant cost impact.

Therefore, GNSS signal design cannot simply follow the “release first, update later” route. It must be fully studied and demonstrated before system implementation. However, as a significant infrastructure, GNSS has a long development cycle. It may take several years from the time of initial signal design to the full operation of the system. This fact forces signal designers to have sufficient foresight, using existing technology, to enable unknown service requirements over future decades of operation. Therefore, although most of the signals in current GNSS have been defined, the evolution of GNSS signal design will not stop there. The full operation of the current GNSS is the beginning of the design work of the next generation GNSS.

From the reality of GNSS design one can find that the growing expanded applications and refined services prompt the new generation systems to broadcast more signals with more complicated structure, which on the one hand makes more efficient use of limited spectrum resources, already crowded with GNSS signals, but on the other hand makes the spectrum crowding situation even worse. Moreover, such a complicated signal structure may make the realization of signal multiplexing more difficult for satellite payloads. Additionally, the limitation of receiving complexity, the requirement for backward compatibility, as well as the demand of interoperability among systems, add many constraints into the GNSS signal design optimization problem. The immaturity of methodology and the conflict between application expansions and resource scarcity cause the signal design of the next generation GNSS to face a series of challenges. At present, it is necessary to take a hard look at the technical challenges in future GNSS signal design, and look for possible solutions in advance.

Challenges Facing Future GNSS Signal Design
Compared with wireless communication signal design, the major feature of navigation signal design is the pursuit of high accuracy ranging capability. If we compare the wireless communication signal, of which the most concerned targets are the capacity and reliability of data transmission, to a paper envelope, then the satellite navigation signal can be compared to a ruler, since its main concern is the accuracy and robustness of the ranging measurement.

In the design process of this “ruler”, the selection of the carrier frequency, the optimization of the spreading modulation, and the multiplexing of signal components are the top three critical and challenging parts.

Carrier Frequency Selection
The carrier frequency, as is the material of a ruler, determines many of the attributes of a navigation signal, including the propagation characteristics, the cost of transmitting and receiving hardware, signal Doppler shift, and possible interference with other radio systems.

Among all available spectrum resources, the L-band has many advantages for satellite navigation applications, such as good propagation characteristics, moderate antenna size, and relatively small atmosphere influence, and therefore became the preferred frequency band for satellite navigation signals. At present, the vast majority of GNSS signals are gathered in the upper L band (1559 ~ 1610 MHz) and the lower L band (1164 ~ 1300 MHz).

Figure 1 (see inset photo, above right) illustrates spectra of navigation signals of the current and emerging GNSSs in the upper L band. Obviously, it is hard to find any unoccupied contiguous segment in this frequency band. There are only a few scattered available frequency fragments remaining between main lobes of existing signals. Interference arises among different signals. Although it is possible to reduce the spectral overlap of the signal located at the same central frequency to a certain extent by using different subcarriers or adjusting the spreading chip waveforms, it is still becoming increasingly difficult to find a suitable central frequency for a newly added navigation signal in L-band.

Some studies (including Avila-Rodriguez et alia, and Irsigler et alia, Additional Resources) consider the use of higher frequency bands, such as the S-band at 2483.5 to 2500 MHz and the C-band at 5010 to 5030 MHz. However, compared with L-band, the valid frequency spectrum allocated to navigation service in S- and C-bands is more limited, so that signals these bands can support are more limited. In addition, using these bands with higher frequencies will result in greater space transmission loss, greater phase noise, and greater Doppler shift.

Spreading Modulation Design
Spreading modulation is to a satellite navigation what scale is to a ruler. The most direct influence of the spreading modulation design is to adjust the signal spectrum shape in order to distribute the power of the signal to a specific frequency position. Research indicates that spreading modulation directly affects the receiving performance for a signal in thermal noise, interference, as well as multipath environments, and RF compatibility between signals in the same frequency band. Therefore, the optimization of the spreading modulation technique is considered one of the most important ways to realize spectrum compatibility and performance improvement simultaneously.

In the new generation GNSS, there are two new trends emerging in spread modulation design. Firstly, signal power distribution is changing from concentrating near the carrier frequency to a splitting spectrum form, which is represented by binary offset carrier (BOC) modulation. Research indicates that splitting spectrum characteristics result in a spectrum separation from legacy signals located at the same central frequency and a wide root mean square (RMS) bandwidth which results in the potential advantage of improved ranging accuracy and inherent multipath resisting ability. Secondly, in addition to bipolar waveforms, more and more multi-level spread modulation waveforms are emerging, such as that in Composite BOC (CBOC) and Alternative BOC (AltBOC) modulations. Relaxing the constraint of waveform level can provide greater freedom for spread modulation waveform optimization, thus providing more possibilities for improvement of the signal performance (Pratt and Owen, and Zhang et alia 2011, Additional Resources).

However, these two trends bring increased complexity to both the transmitter and the receiver. The larger RMS bandwidth of the splitting spectrum signal requires a higher subcarrier frequency, which means a wider receiver frontend bandwidth and the increment of complexity of the receiver. Although the sophisticated receiving strategy is acceptable for high-end applications such as surveying and mapping, the cost and complexity is hard to justify for low-end consumer electronics devices. A direct way to address this problem is by transmitting multiple signals from the satellite, providing high-end users with a wideband signal that uses a complex chip waveform, while allowing the low-end users to have a simple receiving strategy for a narrowband signal with a simple chip waveform. Unfortunately, the increment in the number of signals at the same frequency degrades the multi-access interference (MAI) between the system and the inter-system signals, which leads to deterioration in the receiving performance. Furthermore, the increase in the number of signals and the complexity of the signal waveform pose a challenge to constant envelope multiplexing, which is another critical part of satellite navigation signal design.

Multiplexing
Signal multiplexing refers to the technique of combining multiple different signal components into one signal over a shared transmitting chain. It is not widely treated in most GNSS interface control documents (ICDs) but is the basis for a variety of PNT services for today and the future. Due to the limitation in GNSS transmitting power and the nonlinearity of the amplifier, multiple spreading signals should share a carrier frequency and multiplex into a composite signal with a constant envelope in the signal transmitter.

In the constant envelope multiplexing on the navigation satellite, as the number of multiplexing signals increases, more inter-modulation terms power should be added to keep the envelope of multiplexed signal constant. Since the information carried on inter-modulation terms is redundant for a receiver, the higher proportion of inter-modulation terms mean the less useful power output, which is expressed as a lower multiplexing efficiency, thus reducing the received carrier to noise ratio (CNR). Though increasing the transmitting power can compensate for the multiplexing loss, it will further deteriorate MAI between the system and the inter-system signals.

A Gordian Knot
Under the conventional idea of separately optimizing carrier frequency, spreading modulation, and multiplexing, a Gordian knot is emerging in future GNSS signal design. As shown in Figure 2, in the independent design of these three key elements, it is difficult to reconcile the contradictions among service diversity, ranging accuracy, receiving complexity, radio frequency (RF) compatibility and multiplexing efficiency.

Users always want signal ranging performance to be as high as possible while the receiving complexity is as low as possible. However, one cannot have both at the same time. The most straightforward way to cope with the high-performance demand is to increase the signal bandwidth, and moving the main spectrum component of signal away from the carrier frequency. However, since there are almost no contiguous segments of unoccupied spectrum remaining in the upper L-band, increasing the signal bandwidth will aggravate spectral interference between the new signal and existing signals. Furthermore, a wider signal bandwidth and complex subcarrier structure also result in a higher processing burden on receivers, which is unacceptable by low-end users. The direct way to further support low-end users is to add more narrowband signals with simple structures. Nevertheless, increasing signal numbers not only further increases spectrum interference, but also further reduces the power efficiency of multiplexing. That means the useful signal power is reduced, and also that the interference is increased. As a result, although the original intention is to improve the overall performance, the actual effect is to degrade the performance of each signal component.

In order to get out of the cycle of contradictions among measurement accuracy, services variety, RF compatibility, as well as multiplexing efficiency in satellite navigation signal design, it is necessary to break the routine.

Our inspiration is attributed to Vernier caliper, which combines two rulers with different scales together. In principle, each scale can be used independently as a simple ruler. However, based on the difference between scale divisions of these two rulers, when we use these two scales jointly in a proper way, we can obtain a higher measurement accuracy. Along the way, in navigation signal design, when we re-examine carrier frequency, spreading modulation, and multiplexing these three key elements as a whole, we find a solution to cut the Gordian knot: the multicarrier constant-envelope composite (MCC) signal.

Multicarrier Constant-Envelope Composite Signal
The concept of multi-carrier signals originates from the field of wireless communications. Typical multicarrier communication signals include multi-tone signals, orthogonal frequency division multiplexing (OFDM) signals, and multicarrier code division multiple access (MC-CDMA) signals. However, the satellite navigation signal has two major differences compared with the communication signal: First, the core mission of the navigation signal is the high accuracy time of arrival (TOA) estimation but not the data transmission. The TOA estimation processes based on multi-carrier communication signals, such as OFDM signal, are complex (See Thevenon et alia, Additional Resources) and the inherent high RMS bandwidth performance advantage of multi-carrier signals is difficult to be adequately brought into play. Second, the vast majority of existing multi-carrier signals have a high peak-to-average ratio (PAR), which hinders their application for satellite transmission (See Mateu et alia, and Emmanuele et alia, Additional Resources).

Unlike the above-mentioned multicarrier communication signals, as shown in Figure 3, the proposed MCC signal is like a “spectral transparent adhesive”. It “glues” a plurality of narrowband signal components located in multiple spectral gaps together to form a wideband constant envelope signal, sharing a common up-converter, amplifier chain and antenna aperture. The core features of the MCC signal are:

• Sparsity in the frequency domain;
• Envelope constancy in the time domain; 
• High flexibility on design elements such as the number of sub-bands, sub-band frequency spacing, the number of signal components in each sub-band, shape of the spreading waveform, the power ratio and relative phase of signal components; 
• Transparency of the compositing to the receiver.

Compared with existing solutions, the MCC signal has the following unique advantages:

On the one hand, multicarrier is one of the most effective ways to utilize spectrum gap resources. As previously mentioned, in the increasingly crowded satellite navigation band, the absolute bandwidth of the newly added signal is severely limited. There are only some scattered frequency fragments available between main lobes of existing signals, as illustrated in Figure 3(a). However, the frequency difference between signal components in different sub-bands of the multi-carrier signal can transform this unfavorable factor into a favorable one. As illustrated in Figure 3(b) and (c), placing multiple sets of narrowband signals components in the fragment band gaps and combining them into a wideband constant envelope signal can construct a MCC signal. The spectrum sparse characteristic of such signal can not only ensure adequate spectral separation with existing signals in the same band, but also provide a large RMS bandwidth for better ranging performance, resulting in solving the contradiction between spectral efficiency and ranging performance.

On the other hand, the different narrowband components in the MCC signal can be optimized for targeted PNT services, with different spreading sequences, different spreading waveforms, different power allocations, and different data message structures and contents, to meet future diversified PNT requirements. At the same time, in MCC signals, those components are combined into a whole signal by constant envelope multiplexing that is “transparent” enough for receivers, not only allowing narrowband receivers to process each component separately with low-complexity, but also allowing wideband receivers to process the total or partial components of this signal with a wide RMS bandwidth. That is, the MCC signal has innate features of diversified receiving and processing strategies, which addresses the contradiction between power efficiency and service diversity.

In addition, the integrated structure of the MCC signal ensures a strong correlation between the transmission channel effects of each sub-band component, which creates conditions for joint processing of components in multi sub-bands, such as joint acquisition, joint tracking, and joint pseudorange extraction.

Given the above, a MCC signal can not only achieve outstanding ranging accuracy without significantly increasing the RF interference to the existing signals in the same band, but also provide users with diversified and targeted service without noticeably deteriorating the multiplexing efficiency onboard the satellite. It provides a promising technique solution for the next generation GNSS signal design.

Construction of a MCC Signal Based on the CEMIC Method
The key to the MCC is determining how to combine several flexible signals located at multiple different central frequencies with arbitrary power, chip rate, and spreading waveform into an integral signal with a constant envelope. In the field of satellite navigation signal design, the study of constant envelope multiplexing has long focused at single-frequency cases. Although there are some dual-frequency constant envelope multiplexing methods, such as those described by Lestarquit et alia, Yao et alia 2016 and Zhang in Additional Resources, few of them can support multiplexing for more than two sub-bands. Moreover, the vast majority of existing constant envelope multiplexing techniques are only applicable under strict pre-conditions in the component waveform shape, component number, component power ratio and phase relationship, which is not suitable for the proposed conception of using multiple spectral gaps to carry diversified services.

The recent emergence of a high efficiency generalized multicarrier joint CEM technique for multilevel spreading signals, termed CEM via intermodulation construction (CEMIC), as described by Yao et alia 2017a in Additional Resources, presents the possibility for the realization of a MCC signal. Compared with existing CEM techniques, CEMIC has a much higher design flexibility in the number of sub-bands, the number of signal components, power ratio and phase relationship among components, and the shape of spreading chip waveforms. CEMIC can be applied to any number of bipolar or multilevel spreading spectrum signals with arbitrary power distribution at one or more subcarrier frequencies. Such a high degree of design flexibility provides system designers great room in signal scheme optimization for varied navigation applications in the future.

In this section, based on the design theories presented in Yao et alia 2017a, and Yao et alia 2017b, Additional Resources, an implementation technique of MCC with extremely high design flexibility is presented.

Consider combining N spreading spectrum signal components located at several sub-bands, s_i(t), for i = 1 = N , into a composite signal with constant envelope. In principle, there is no constraint on the spreading code rate and spreading waveform shape for each s_i(t). However, for simplicity, here we assume that all of the s_i(t) are MCS signals with the same code rate T_c, and every MCS symbol is divided into M segments with equal length T_s = T_c/M. More general cases can be found again in Yao et alia 2017. Then can be mathematically expressed as

Equation (1) (see inset photo, above righ, for all equations)

where c_i is the navigation data modulated by the corresponding spreading code, p_i is the waveform value in k-th segment, and ψ(t) is unit amplitude rectangular pulse function with T_s duration.

Define f_i as the frequency offset of the subcarrier of s_i(t) from the carrier frequency f₀, the selection of which depends on the specific location of spectral gaps. For the convenience of digital implementation, the subcarrier waveform can choose the sample-and-hold version of the complex sinusoid waveform

Equation (2)

where Δf_i = f_iT_s. Since for all of the deployed GNSS signals, the carrier frequency, the spreading chip rate, as well as the subcarrier rate are all the multiple of 1.023 megahertz, by choosing the proper values of T_c, f_i, and M, it is easy to ensure that ΔT_i = 1 / Δf_i is an integer. That means the signal component modulated by the subcarrier is still a MCS signal. 

Directly combining these N signal components into a multicarrier composite signal is mathematically equivalent to constructing a new baseband signal, of which the complex envelope is

Equation (3)

where c̃_i[m] = c_i[n]p_i[k] e^j2πΔf_im with n = [m T_s/T_c] and, k = [(mT_s − nT_c)/T_s], P_i and θ_i are the power allocation and initial phase of i-th component respectively.

It can be seen from the above equation that s_MUX(t) is also a MCS signal with segment length T_s, and in every duration t[m] ∈ [mT_s,(m+1)T_s) , its value is fixed to

Equation (4) 

In general, the MCS waveform may have multi amplitude levels. For simplicity, assume that every p_i has up to K different possible values, while every sample-and-hold version subcarrier waveform has up to ΔT_i different values. Then s_MUX(t) has a maximum of

Equation (5)

possible complex values, resulting in the temporal fluctuation of the envelope. In order to keep the envelope of the multicarrier composite signal constant, the basic idea of CEMIC is adding an additional component I_IM(t) to s_MUX(t) to compensate for the envelope fluctuation. This additional component can be referred to as the intermodulation (IM) term. In every time period t[m] ∈ [mT_s, (m+1)T_s], the value of I_IM(t) is determined by the value of s_MUX(t[m]), to ensure the composite signal s_CE(t) = s_MUX(t) + I_IM(t) is a constant envelope signal. As pointed out in Yao et alia 2012, this is equivalent to finding an amplitude mapping rule f: {c̃₁, c̃₂, …, c̃_N} ↦ I_IM, that gives values to the IM term for each of the F combinations of values of c̃_i, to make the envelope of the superposed signal be constant. 

If only the envelope constancy of s_CE is constrained, then an infinite number of mapping rules can be used. However, since the IM term I_IM(t) is only used to maintain the constancy of signal envelope, from a receiving power efficiency standpoint, its proportion should be as low as possible in the whole composite signal, and its influence on receiving performance should be as small as possible. The core of CEMIC is to find an optimal mapping rule, which constructs an IM term that can guarantee the optimal power efficiency, minimal impact on the correlation characteristics of useful components, and the envelope constancy of the composite signal.

Generally, using CEMIC to construct a MCC signal has the following four main steps:

1) According to T_c, f_i, M, and the shapes of p_i, for i = 1 = N, list all F possible combinations of values of {c̃₁, c̃₂, …, c̃_N}, and construct the component weight vectors

c̃_i = [c̃_i⁽¹⁾, c̃_i⁽²⁾, …, c̃_i^(F)]^T     (6)

for i = 1 = N, where c̃_i^(ℓ) is the value of c̃_i in the ℓ-th combination. Note that, in order to distinguish from the time index that is in square brackets, here the combination index is put in parentheses. 

2) Based on component weight vectors c̃₁, c̃₂, …, c̃_N using the Gram-Schmidt orthogonalizing method or other methods, construct a set of orthogonal vectors {g₁, g₂, …, g_F-N} to make span{g₁, g₂, …, g_F-N} be the orthogonal complement space of {c̃₁, c̃₂, …, c̃_N}. A general construction algorithm for this step is given in Appendix A of Yao et alia 2017. However, if all of p_i are bipolar, a much simpler direct construction method proposed in Zhang et alia 2012, Additional Resources can also be available.

Define
s_CE = Cr + Gw, where C = [c̃₁, c̃₂, …, c̃_N], G = [g₁, g₂, …, g_F-N], and r = [√P₁e^jθ₁, √P₂e^jθ₂, …, √P_Ne^jθ_N]^T and solve the following constraint minimization problem

Equation (7) 

to obtain the optimal coefficient vector w, where s^(ℓ)_CE is the ℓ-th entry of s_CE. 

Let λ = Gw = [λ⁽¹⁾, λ⁽²⁾, …, λ^(F)]^T. Then we obtain the optimal mapping rule from the value combination of N signal components to the IM term I_IM(t). In every moment, if the values of {c̃₁, c̃₂, …, c̃_N(t)} correspond to the ℓ-th value combination, I_IM(t) takes the value λ^(F), and s_CE(t) = s_MUX(t) + I_IM(t). 

Case Study of Adding a MCC Signal in L1 Band
As a sample application, consider adding a new MCC signal in the L1 band. Although current GNSSs do not have such a plan yet, through this specific example, one can clearly see the design process of a MCC signal, and the characteristics and advantages of this signal in both the transmitter and receiver.

As mentioned, the upper L-band has been overcrowded. All GNSSs broadcast their open and authorized service signals in this band. If a new wideband signal is added to this band, it will be hard to avoid significant spectrum overlapping with the existing signals. However, it is noted that most of the existing signals in this band have spectral nulls at 1575.42, ±4.092, ±8.184, and ±10.23 megahertz, etc. Thus, under the premise of ensuring good RF compatibility, a possible new signal solution is to place multiple narrowband components in these spectral gaps.

For simplicity, consider the case of multiplexing five narrowband components with BPSK-R(0.5) spreading modulation in this example. As shown in Figure 4, the center frequencies of these five components are set to 1577.466 MHz, 1579.512 MHz, 1581.558 MHz, 1583.604 MHz, and 1585.65 MHz, respectively. In the transmitter, the carrier frequency of the composite MCC signal can be 1581.558 MHz. Thus, the subcarrier frequencies of components are f₁ = –4.092 MHz, f₂ = –2.046 MHz, f₃ = 0, f₄ = 2.046 MHz, and f₅ = 4.092 MHz, respectively. Under the equal power assumption, r can be set to [1,1,1,1,1]^T. In this design case, considering the constraint of implementation complexity, M should not be too large. Here we take M = 8, the shortest segment length T_s = (32 °— 1.023e6)^–1 s.

The theoretical power spectral density (PSD) of these five narrowband components before the constant envelope reconstruction is shown in Figure 5(a). Following four steps of the CEMIC method presented in the previous section, the MCC signal is constructed, for which the theoretical and simulation PSDs are shown in Figure 5(b) and (c), respectively.

By comparing Figure 5(a) and (b), it is observed that after constant envelope reconstruction, the PSD of the MCC signal is different from that of the direct superposition of these five components. The difference is mainly reflected in the appearance of the IM term in MCC signal. It can be seen that the power of the IM term is much lower than that of the useful signal components, and the difference is at least 10 decibels. In fact, the multiplexing efficiency of this example is 80.41%. That is, the newly added IM term accounts for only 19.6% of the total signal power, and its spectrum is distributed far away from the carrier center frequency.

Figure 6 shows the modulation constellation of the MCC signal. As illustrated in the figure, all the phase points are distributed on a circle. This feature enables the payload high power amplifier (HPA) to operate in its full-saturation mode to maximize power conversion efficiency.

RF Compatibility Analysis
To evaluate the RF compatibility between the newly added MCC signal and the existing BPSK-R(1), MBOC(6,1,1/11), BPSK (10), as well as BOC (10,5) signals in the same frequency band, we calculate their spectral separation coefficient (SSC). The receiver front-end filter is assumed to center on 1575.42 MHz, with a single side bandwidth of 12 megahertz, which is sufficient to cover the highest frequency component of the MCC signal. Table 1 shows the SSC of the existing signals with MCC signal.

It can be seen that the MCC signal maintains good RF compatibility with the existing signals by effectively utilizing the fragment band gaps between the main lobes of existing signal spectrum. It can be verified that if more sub-bands are employed, or moving f₅ from 1575.42 + 10.23 MHz to 1575.42 + 12.276 MHz, the RF compatibility between MCC signal and the BOC(10,5) signal can be further improved.

Diversified Processing Strategies
As previously mentioned, in addition to the effective utilization of the spectrum resource, another key advantage of the MCC signal is that it inherently has multiple receive modes, providing a variety of processing strategies for receivers with different performance and complexity constraints.

Since the MCC signal is composited in the digital baseband, the subcarrier phase of each component is completely coherent, and components within the MCC signal pass through the same transmission channel, the errors introduced by thermal noise, multipath, as well as the dynamic stress also have strong correlation. The receiver can either treat these signal components separately, or jointly process multiple components or even the entire composite signal as a whole.

The simplest processing mode is treating the narrowband components in the MCC signal as different signals. Such a processing mode requires minimal processing complexity. If narrowband components employ BPSK-R spreading modulation, as discussed in this example, their acquisition and tracking methods can be directly inherited from the traditional cases, where both the rectangular pulse spreading chip and the sinusoidal subcarrier waveforms can be employed in the local replica. The cross-correlation function (CCF) between the received MCC signal and the local replica of each signal component in this processing mode is shown in Figure 7(a).

If the receiver jointly processes three signal components, which are s₂(t), s₃(t), and s₄(t), without loss of generality, the local replica can be

s_local(t) = s₂(t)e^j2Πf₂t + s₃(t) + s₄(t)e^j2Πf₄t     (8)

The CCF between the received MCC signal and this local replica is shown in Figure7(b). 

Further, the whole MCC signal can even be used as the local replica to realize the matching receiving, for which the CCF is shown as Figure 7(c).

In order to quantitatively compare the performance of above three processing modes, equivalent Gabor bandwidth, correlation loss, and average multipath error envelope are used as evaluation indexes to measure the code tracking accuracy, the performance of acquisition and demodulation, and the multipath resisting performance, respectively.

Figure 8(a) and (b) show the equivalent Gabor bandwidth and the correlation loss of these three processing strategies with respect to the front-end double-sided bandwidth, respectively. Figure 8(c) shows the average multipath error envelopes of these three processing strategies, with front-end double-sided bandwidth of 10 megahertz, and multipath-to-direct ratio (MDR) of –5dB.

One can see from Figure 8 that in different processing strategies, the receiving performance presents an obvious graded characteristic. With a narrow bandwidth, the single-component processing mode has the minimum processing complexity, but the largest correlation loss and the lowest ranging accuracy. However, as an increasing number of components are processed jointly, for wideband receivers, not only is the correlation power loss decreased, but also a higher ranging accuracy as well as a better multipath resisting ability can be obtained. That means the innate multiple processing strategies of the MCC signal can provide different tradeoffs between performance and processing complexity to different PNT application requirements. With MCC signals, receivers can obtain various levels of receiving performance by jointly processing different subsets of signal components. This is one of the major advantages of the MCC signal.

Processing Mode Switching
The inherent multi-strategy processing advantage of MCC signal can be taken not only by different types of receivers, but also in different stages of a wideband receiver. The MCC signal allows the receiver to dynamically switch the processing strategy at different processing stages, according to the current working status to achieve balance between processing complexity and accuracy.

From Figure 7, it can be seen that the main peak of CCF under the single-component processing mode is the widest, which can widen the acquisition bins and provide an unambiguous large pull-in range to the tracking loop. As more components are jointly processed, the energy of the CCF increases significantly, and the main peak of the CCF becomes sharper, which implies higher potential tracking accuracy. However, more side peaks appear on both sides of the CCF main peak.

One possible strategy for a wideband MCC signal receiver is using the single-component processing mode in the initial acquisition and pull-in phases, utilizing the wide CCF main peak to obtain a wider search step and a larger pull-in range. After the tracking loop is stabilized, three- and five-component joint processing can be employed incrementally, gradually gaining higher signal-to-noise ratio and sharpening the CCF peak to obtain higher tracking accuracy.

The switching strategies of the processing mode of MCC signals are not limited to this simple mode. In fact, the multi-component multi-subcarrier structure of the MCC signal provides the possibility for the future receivers to explore the diverse switching strategies.

Selective Availability
Since the multiplexing used in MCC signal construction is sufficiently flexible, different signal components in the MCC signal can be configured with different pseudorandom (PN) codes and different spreading modulation waveforms, and can be modulated with different data messages. Therefore, the service provider can assign different codes and messages to different signal components, controlling the access permissions and providing selective performance to different user levels. The receiver selects the corresponding processing mode according to its own privilege level and thus obtains the available acquisition, tracking, and demodulation performances.

For example, as shown in Figure 9, in the five-component design case provided in this section, s₃(t), which is located on the carrier frequency, can be assigned to be the open access signal, of which the PN code generation method and the data message structure are fully open. All receivers can access this component with the single-component processing mode, thus obtaining a relatively low signal-to-noise ratio, and a basic ranging accuracy level.

Components s₂(t) and s₄(t), which are with relatively low subcarrier frequencies, can be assigned as the secondary authorized signals. Their PN code generating information and data message structures are only provided to the authorized secondary users. These users can access three components, so that multi-component joint processing and dynamical mode switching strategies can be used to obtain the improved performance.

Components s₁(t) and s₅(t) can be assigned as senior authorized signals, with encrypted PN codes and data message structures, serving authorized senior users. Senior authorized receivers can access all the five components, to obtain the most diversified processing strategies, the highest signal-to-noise ratio, and the highest ranging performance.

In addition, if the data structures of different components are well-designed to carry complementary messages — for authorized users who can access multiple components — the time to first fix (TTFF) can be effectively reduced. In theory, the TTFF of a receiver that jointly processes five components can be shortened by 80% over that of the basic single-component receiver.

The case study in this section demonstrates that the MCC signal has a high degree of flexibility in both the broadcasting strategy and the receiving strategy. There are many more possible broadcasting and receiving modes of MCC than those discussed in this example. In fact, this signal structure offers a wide design space for both the system providers and the receiver developers in future.

Conclusions
As a significant infrastructure, GNSS has a long development cycle. This characteristic means that we can only employ existing techniques to meet the demands over the next few decades. Although it is impossible to envision GNSS products and services further out in time, we can enable future development by implementing excellent signal designs with higher adaptability and flexibility.

The contradiction between the need for performance improvement and the fact that power and spectrum resources are limited will be more serious in the next generation of GNSS signal designs. In order to solve this contradiction, this article first proposes the concept of a multi-carrier constant envelope signal, and studies its feasibility as the next generation satellite navigation signal. A corresponding design method based on the CEMIC technique is given, and an example is presented to demonstrate the RF compatibility, typical receiving strategies, corresponding performances and selection availability of MCC signals. The analyses show that the MCC signal can make full use of the existing spectrum resources, providing both various broadcast strategies and multiple receiving strategies with a variety of performance levels for different categories of users. This technique can serve as a practical new solution to the next generation satellite navigation signals design.

Acknowledgment
This article is based on a presentation given by the first author at the ION GNSS+ 2017 conference on September 27-29, 2017, hosted by the Institute of Navigation in Portland, Oregon, USA. This work is supported by National Natural Science Foundation of China (NSFC), under Grant 61771272.

Additional Resources
[1] Avila-Rodriguez, J.-A., S. Wallner, G. W. Hein, B. Eissfeller, M. Irsigler, and J.-L. Issler, “A vision on new frequencies, signals and concepts for future GNSS systems,” in Proceedings of the 20th International Technical Meeting of the Satellite Division of The Institute of Navigation, Fort Worth, TX, 2007, pp. 517-534.
[2] Emmanuele, A., et alia., “Evaluation of Filtered Multitone (FMT) Technology for Future Satellite Navigation Use,” Proceedings of the 24th International Technical Meeting of the Satellite Division of the Institute of Navigation (ION GNSS 2011), pp. 3743-3755, 2011.
[3] Irsigler, M., G. W. Hein, and A. Schmitz-Peiffer, “Use of C-Band frequencies for satellite navigation: benefits and drawbacks,” GPS Solutions, vol. 8, no. 3, pp. 119-139, 2004.
[4] Lestarquit, L., G. Artaud, and J.-L. Issler, “AltBOC for Dummies or Everything You Always Wanted to Know About AltBOC,” presented at the ION GNSS 2008, Savannah, GA, US, 2008.
[5] Mateu, I., et alia., “A search for spectrum: GNSS signals in S-band, part 2,” Inside GNSS, vol. 2010, no. October, pp. 46-53, 2010.
[6] Pratt, A. R. and J. I. Owen, “BOC modulation waveforms,” in Proceedings of the 16th International Technical Meeting of the Satellite Division of the Institute of Navigation, ION-GPS/GNSS-2003, 2003, pp. 1044-1057.
[7] Thevenon, P. et alia, “Pseudo-Range Measurements Using OFDM Channel Estimation,” (in English), Proceedings of the 22nd International Technical Meeting of the Satellite Division of the Institute of Navigation (ION GNSS 2009), pp. 481-493, 2009.
[8] Yao, Z., J. Zhang, and M. Lu (2016), “ACE-BOC: dual-frequency constant envelope multiplexing for satellite navigation,” IEEE Transactions on Aerospace and Electronic Systems, vol. 52, no. 1, pp. 466-485, 2016.
[9] Yao, Z., F. Guo, J. Ma, and M. Lu (2017a), “Orthogonality-based generalized multicarrier constant envelope multiplexing for DSSS signals,” IEEE Transactions on Aerospace and Electronic Systems, vol. 53, no. 4, 2017.
[10] Yao, Z., and L. Mingquan. (2017b) Signal Multiplexing Techniques for GNSS: The principle, progress, and challenges within a uniform framework. IEEE Signal Processing Magazine.
[11] Zhang, X. M., X. Zhang, Z. Yao, and M. Lu (2012), “Implementations of Constant Envelope Multiplexing based on Extended Interplex and Inter-Modulation Construction Method,” (in English), Proceedings of the 25th International Technical Meeting of the Satellite Division of the Institute of Navigation, pp. 893-900, 2012.
[12] Zhang, X., Z. Yao, X. Zhang, and M. Lu (2011), “A Method to Optimize the Spreading Code Chip Waveform in Sense of Gabor Bandwidth,” in Proceedings of the 24th International Technical Meeting of The Satellite Division of the Institute of Navigation (ION GNSS 2011), 2011, pp. 1299-1304.
[13] Zhang, K., “Generalised constant-envelope DualQPSK and AltBOC modulations for modern GNSS signals,” Electronics letters, vol. 49, no. 21, pp. 1335-1337, 2013.

The post Spectral Transparent Adhesive appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

Q: Do modern multi-frequency civil receivers eliminate the ionospheric effect?

Q: Do modern multi-frequency civil receivers eliminate the ionospheric effect?

A: It is common knowledge in the GNSS community that the ionosphere is dispersive in the L-band, meaning the refractive effects on the carrier phases are proportional to the wavelengths of the carriers, in turn causing differential variation in the measured codes and phases of the various navigation signals transmitted by the satellites. Use of multiple signals of distinct center frequency transmitted from the same GNSS satellite allows direct observation and removal of the great majority of the ionospheric delay, and gives the impression to users that the ionosphere may not be a problem for modernized receivers. While the general assumption of nearly perfect correlation between the effects measured on multiple independent signals is correct in normal conditions, it does not appear to hold in the presence of ionospheric scintillation.

High and Low Latitude Scintillation Effects
Scintillation refers to random fluctuations in the received wave field strength (“signal fading”), as well as phase and group delay caused by the irregular structure of the propagation medium. Ionospheric scintillations are random rapid variations in the intensity and phase of the received signals resulting from plasma density irregularities in the ionosphere.

Many of the important contributors to ionospheric scintillation are already known, such as the variation of scintillation activity with magnetic activity, geographic location, local time, season, and the 11-year solar cycle.

The most significant and frequent scintillation activity including both phase and amplitude variations is observed in low latitude regions within about 15° of the Earth’s magnetic equator, particularly in the hours after local sunset. In high latitude regions scintillation is frequent but generally less severe in terms of signal tracking disruptions than that in the equatorial regions. The high-latitude environment can be divided into two subregions, the polar caps (regions around the magnetic poles) and the auroral zones (approximately circular regions around the two geomagnetic poles located at about 67° north and south geographic latitudes, and about 3° to 6° wide). Of these, the polar cap experiences both amplitude as well as phase scintillation activity, while mainly phase scintillation is observed at high latitude auroral regions.

In mid-latitude regions scintillation is rarely observed, but during intense ionospheric storm conditions phenomena can extend into the mid-latitudes.

Figures 1 and 2 (see inset photo, above right, for all figures) show examples of ionospheric scintillation as observed on the detrended signal intensity (effectively power) and detrended carrier phase measurements at 69.5° latitude (Tromsø, Norway) and 21° latitude (Hanoi, Vietnam), respectively. In the particular event shown in Figure 2 the depth of fades reaches 43 decibels (dB) on L1CA which is severe by any metric, and is a substantial qualitative difference from the high-latitude phase scintillation events where only very weak fading activity is typically observed. It should be noted that ionospheric activity is more dependent on the geomagnetic latitude of the user than the geographic latitude. While it might be clear that Tromsø station is located in the auroral region, the Hanoi station has somewhat lower geomagnetic latitude than geographic latitude and is in fact located within the equatorial zone.

Since ionospheric scintillation is essentially a rapid variation in the apparent ionosphere it is easy to assume that the typical approaches applied for removing ionospheric influence will be effective during scintillation.

Ionosphere-Free Combination
The advantage of multi-frequency GNSS receivers in terms of handling the ionospheric error is that they can combine carrier phase measurements at different frequencies to cancel out the first order effect due to ionospheric refraction. The receiver does not typically measure the ionospheric delay directly, but is using the so-called ionosphere-free linear combination of the observables. Consider generalized versions of carrier phase measurements on two frequencies, i and j, expressed in meters:

Φ_Li = ρ + λ_iN_i − I_i     (1)
Φ_Lj = ρ + λ_jN_j − I_j

where ρ is the geometric range between the satellite and the receiver; λ_i and λ_j are the wavelengths, N_i and N_j are the integer ambiguity terms, I_i and I_j are the ionospheric propagation delay errors. For simplicity, the receiver noise and multipath errors are not included. The expression for an arbitrary linear combination of two carrier phase measurements can be written as follows: (For more on this topic, read the GNSS Solutions column from January/February 2009).

Φ_ij = αΦ_Li + βΦ_Lj     (2)

where α and β are constants. This allows one to model a linear combination of phases in the same way as the individual observables:

Φ_ij = ρ + λ_ijN_ij − I_ijη     (3)

In (3), λ_ij is the wavelength, N_ij is the integer ambiguity term, and I_ij is the ionospheric propagation delay error for the linear combination. In order to remove the ionospheric error (η = 0), but leave the geometric portion unchanged and the resulting ambiguity still an integer, the ionosphere-free combination has been proposed:

Φ_IFree = f_i²Φ_Li − f_j²Φ_Lj  ∕ f_i² − f_j²     (4)

where f_i and f_j are the carrier frequencies expressed in hertz. The phase scintillation is, however, caused by both refractive and diffractive effects. The diffractive effects cause rapid transitions in the phase which do not scale with the carrier wavelength resulting in a residual error in the ionosphere-free linear combination (4) of phase measurements.

While this correction term is for most purposes considered complete, there are factors that can cause apparent deviation between the two carriers including multipath, receiver noise, and un-modelled terms in (4). Corrections produced using (4) will have a residual error due to second and third order dispersion effects, which are conservatively bounded to 0–2 centimeters and 0–2 millimeters at zenith respectively, under an assumption of a 100 TECU (total electron content unit; 1 TECU ≈ 16 cm at GPS L1) background ionosphere. Since 100 TECU is a high value for zenith ionosphere the value of the higher order terms will often be well below 2 centimeters instantaneously, and will vary by only a small fraction of this amount over short time periods.

Although some recent findings have shown that magnitudes of 3 centimeters referenced to L1 are possible due to the higher order terms, it has also been shown that the variation rate is typically limited to the level of centimeters per hour.

During phase scintillation events it is possible that the multiple carriers of a given satellite will (when scaled for frequency as in Figure 3) track each other within the margins of error expected when accounting for thermal and oscillator phase noise on each channel. However, it is also possible that near total de-correlation of the phases will occur during phase scintillation accompanied by fading events as is depicted in Figure 4 where the detrended scaled carrier phase observables from L1, L2 and L5 transmitted by a block IIF GPS satellite visibly deviate from one another. Even the closely-spaced L2 and L5 carriers exhibit substantial decorrelation, equivalent at times to a full L1 carrier cycle of nearly 20 centimeters, well outside of the level which could be plausibly attributed to higher order terms ignored by (4).

On close inspection, the data shown in Figure 4 does not appear to contain any stepwise transitions of a magnitude commensurate with a full or half cycle slip on any of the carriers, meaning that this decorrelation is unlikely to be a signal tracking error. Unlike static group delay errors, it is not possible to measure and estimate this error contribution a priori. It is effectively an additional noise source present only during scintillation. Since it will influence the magnitude of the residual error in the case of multi/dual frequency processing it is interesting to analyze this phenomena and attempt to quantify its expected magnitude by considering the level of correlation between carriers during a cross section of scintillation events affecting modernized civil signals believed to be free of cycle slips.

To quantify the correlation level between the scintillation effects on GNSS frequencies, the phase correlation coefficient can be calculated for the observed scintillation events according to the following relationship:

ρ_δφ = ⟨δφ₁δφ₂⟩/(⟨δφ₁²δφ₂²⟩)^1/2, −1 < ρ_δφ < 1     (5)

where the terms δφ₁ and δφ₂ represent epoch to epoch changes in the detrended phases. Figures 5 and 6 show the results for the events observed at 69.5° latitude (Tromsø, Norway) and 21° latitude (Hanoi, Vietnam). In Figure 6 the level of correlation versus the intensity of the phase variation is plotted for L1CA vs. L2CM, and in contrast to the high latitude example shown in Figure 5, where increasing phase instability leads to an increasing level of phase correlation between the two carriers, for the Hanoi data the outcome is entirely different. Indeed, the phase correlation between the two carriers appears to be nearly non-existent on average, as the distribution of correlation measures is bifurcated with half the distribution tending towards higher positive correlation levels, while the other half of the sampled distribution tends towards anti-correlated results.

Ionosphere-free Residual
When scaled by their wavelengths, the carrier phase measurements on different GNSS frequencies appear to match closely when the scintillation effect is weak or moderate, but diverge from one another when the scintillation effect is strong regardless of whether the dominant scintillation effect is on the phase or amplitude of the signal. It is believed that these divergences occur when diffraction alters the phases by a factor that is not proportional to the wavelengths of their carriers leading to a residual in the ionosphere-free phase combination. An example of this phenomena is the so-called “canonical fade”, which may be the cause of the decorrelation events presented here. Figure 7 shows the average absolute L1/L2 ionosphere-free combination residual from Tromsø (69.5° N) observations, whereas results generated based on the data from Hanoi (21° N) are illustrated in Figures 8, 9 and 10.

Noting that the range of carrier phase standard deviation considered in Figures 8, 9 and 10 is smaller than that considered in the high latitude plot, it is clear that the level of ionosphere-free residual present in the Hanoi data increases much more rapidly with rising phase standard deviation than was the case with the high latitude observations.

While it is not unexpected that the L1/L5 combination residual is also substantial, as indicated in Figure 9, the more interesting observation is that the L2C and L5 signals also have considerable levels of decorrelation despite their relatively small 51 megahertz of spectral separation, compared to the nearly 350 megahertz of spectral separation between L1 and L2. In Figure 10, it is seen that for one of the tracked satellites during this event, the level of ionosphere-free residual in the L2CM/L5Q combination seems to exceed one meter even while the underlying data shows no signs of cycle slips.

Conclusion
To summarize, it has been demonstrated that ionospheric scintillation phenomena tend to cause an additional measurement residual in the nominal ionosphere-free combinations that greatly exceeds the expected value of the neglected higher order terms and may be a substantial or dominant nuisance term in some applications. While the residual is present with both phase and amplitude scintillation, and is more pronounced with strong scintillation to a point, the relationship appears stochastic and not deterministic.

It is tempting to assume that concerns about ionospheric effects during all but deep amplitude fades would disappear when users had switched from semi-codeless multi-frequency observables to the use of modernized civil signals due to their much higher tracking robustness. Instead, it seems that even with the modernized signals there is a measurable and occasionally meter level sense in which the ionosphere-free observables are not at all free of ionospheric influence.

Additional Resources

For additional information about ionospheric scintillation:
[1] Conker, R.S., M.B. El-Arini, C.J. Hegarty and T. Hsiao (2003) “Modelling the effects of ionospheric scintillation on GPS/Satellite-Based Augmentation System Availability”, Radio Science, vol.38, no.1.
[2] Carrano, C. S., K. M. Groves, W. J. McNeil, and P. H. Doherty (2013), Direct measurement of the residual in the ionosphere-free linear combination during scintillation, Proceedings of the 2013 Institute of Navigation ION NTM meeting, San Diego, CA, January 28-30, 2013.

For additional information on higher-order ionospheric effects:
[3] Hoque, M.M., and N. Jankowski (2007), “Higher order ionospheric effects in precise GNSS positioning,” Journal of Geodesy number 81, pp 259-268
[4] Liu, Z., Y. Li, J. Guo, and F. Li (2016) “Influence of higher-order ionospheric delay correction on GPS precise orbit determination and precise positioning,” Geodesy and Geodynamics, Volume 7, Issue 5, September 2016, pp 369-376

For information about canonical fades:
[5] Liu, Z., Y. Li, J. Guo, and F. Li (2016) “Influence of higher-order ionospheric delay correction on GPS precise orbit determination and precise positioning,” Geodesy and Geodynamics, Volume 7, Issue 5, September 2016, pp 369-376.

The post Do modern multi-frequency civil receivers eliminate the ionospheric effect? appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

GNSS SDRs are a rapidly advancing area in GNSS receiver research and design. The last few years have brought tremendous growth in this field. Universities and other research institutions have developed and demonstrated advanced capabilities, particularly with respect to multi-constellation GNSS and GNSS-plus-multi-sensor navigation processing for challenging environments. The recent commercial availability of multi-sensor data collection equipment, development platforms and open-source software projects have catalyzed this rapid pace of innovation. Indeed, SDR will likely become a significant commercial GNSS receiver architecture by the end of this decade due to today’s ongoing deployment of multiple global and regional satnav constellations with their diverse signal structures, and rapid advancements in parallel low-power processors and inexpensive sensors.

Non-realtime SDR operational scenarios involve storage and post-processing of samples. These sampled data files can also be used in radio frequency (RF) playback systems for GNSS receiver test and evaluation in the lab. Post-processing and/or playback requires several generic front-end parameters such as RF and intermediate frequency (IF) center frequencies, sample rate, file format, as well as GNSS-specific information such as antenna location and type. We define this information as GNSS SDR metadata. Manual transfer of metadata is the method predominantly used today – a process that is cumbersome and error prone to say the least. No established method exists to exchange this metadata automatically.

During the ION GNSS+ 2013 conference, a group of attendees discussed the need for a formal standard for the exchange of GNSS SDR metadata. This group identified the following benefits of engaging in this activity at the present:

• It identifies and brings the international GNSS SDR community together as a working group. This collaboration is critical in order to get broad acceptance and usage.
• Standardization will help to avoid technology segmentation issues while promoting the pace of innovation by standard practices and compliant tools. 
• The formal standard, if widely adopted, will help ensure compatibility and interoperability of future GNSS SDRs. Specifically, front-end agnostic “plug-and-play” SDRs are envisioned. These have the potential for revolutionizing positioning, navigation, and timing (PNT) systems of the future.

Most authors of significant GNSS SDR publications and contributions over the past two decades are ION members and frequent meeting attendees. Hence, we decided to pursue this standard through ION sponsorship. During the January 2014 Council Meeting in San Diego, ION approved the process for establishing a formal standard. The ION GNSS SDR Metadata Working Group (WG) was formed in April 2014. Membership represents academia, industry (including GNSS SDR product vendors as well as traditional GNSS equipment manufacturers), non-profit research entities, and government agencies spanning countries in America, Europe, Asia and Australia.

Justification for GNSS Metadata Standardization
Figure 1 (see inset photo, above right, for all figures) shows the metadata transfer scheme mainly used in today’s GNSS SDR systems. The top row depicts data collection system (DCS) A, producing an SDR file of format A, consumed by SDR processor A. This may represent an end-to-end solution provided by a vendor or a system that was initially developed around a specific hardware platform. In any case, assume that the metadata for decoding Format A is hard-coded into the processor. Consequently, supporting other file formats involves extensive software revisions on a case-by-case basis. One group may want to share SDR files from DCS A with other groups using SDRs X and Y for research collaboration. This involves conveying the file format and other relevant information accurately to these other groups. Today, this metadata transfer occurs in an ad-hoc way that is prone to interpretation errors.

The second row depicts multi-stream DCS B that produces files with a more complicated format. SDR processors X and Y represents more flexible SDRs that are able to support multiple file formats. However, the data/metadata association requires manual intervention.

DCS C multiplexes other data (such as sensor data) along with multiple GNSS sample streams in the same file. This type of multiplexed collection has the potential to become more common with emerging SDR-related data streaming standards such as VITA-49 (See T. Cooklev et alia, Additional Resources). In this case, custom-designed processor C represents an SDR that fully supports multi-sensor integration capabilities. Because DCS C’s GNSS stream parameters are open, SDR Y is also able to support GNSS-only processing using an ad-hoc metadata transfer scheme. The sensor data parameters in Format C may or may not be open.

As clear from Figure 1, today’s ad hoc methods of metadata exchange do not encourage interoperability and instead cultivates potential for technology segmentation (i.e., various groups developing their own stove-piped solutions and technologies).

Figure 2 shows the same systems of Figure 1 that have adopted a metadata standard. As shown, each DCS produces a compliant metadata file along with the SDR file. The metadata file is read-in by the compliant SDR processor to correctly decode and process files seamlessly.

Adoption of a metadata standard benefits DCS developers because their systems become applicable to a much wider group of users. Similarly, an SDR processor’s utility is extended when it is capable of supporting many file formats from multiple sources seamlessly. Thus, metadata standardization promotes interoperability of GNSS SDR systems and greatly simplifies the exchange of files between groups.

Metadata standardization also benefits other use cases beyond post-processing GNSS SDRs. For example, consider the use of the metadata specification to synthesize compliant SDR files for use in RF playback systems. Additionally, libraries of compliant SDR file sets containing various real-world scenarios could be used interchangeably in compliant RF playback simulators for repeatable and consistent testing of GNSS receivers.

SDR Data Collection Topologies
The ION Executive Committee stipulated that this standardization activity shall not create an unfair advantage or disadvantage to any entity. Specifically, the standard shall not require any existing system to undergo data format changes to achieve compliance. This “do no harm” stipulation implies that the standard be designed such that it supports all current and future SDR file formats. This also means that the WG must “get it right the first time” since major revisions to the standard would be undesirable and adverse to the goal of widespread adoption. Hence, we considered the entire space of possible GNSS SDR data collection topologies. Figure 3 illustrates these topologies.

Figure 3.a illustrates the simplest data collection topology that can exist. This is when a single swath of RF spectrum (referenced henceforth as a “band”) is down-converted and sampled to produce a single data stream. The stream, which may be IF sampled (real samples) or baseband sampled (complex samples) is written to disk as a single file.

Figure 3b represents a DCS that writes parts of a single data stream across multiple files. This may be done to reduce the sustained write performance requirement of storage drives (similar to striping mode in RAID systems).

Figure 3.c is similar to Figure 3.a, except that the data stream represents more than one RF band. An example of this topology is a direct RF sampling front-end architecture that intentionally aliases multiple bands to fall next to each other at baseband. Some bands may be spectrally inverted as a result of the digital down-conversion process.

A DCS may produce multiple data streams. Each stream may contain information from one of many antenna elements (as shown in Figure 3.d, where each stream may also encompass multiple bands as in Figure 3.c). Alternatively, a stream may be sampling one of several bands received by a single wideband antenna, where the multiplexed streams written to file represents channelized samples. Combinations of the above are also possible.

Each stream may also be sampled at different rates and bit depths. For example, consider a civilian GPS L1, L2, L5 system. In this case, the L1 and L2 streams may be sampled at rate f₀ and the L5 stream at 10·f₀ (since the L5 signal’s null-to-null bandwidth is 10 times wider than L1 C/A and L2C), where f₀ represents the base sample rate. Figure 3.d may also represent how these multiple streams are multiplexed into a single lane of packed binary data and written to a single file.

Similar to Figure 3.d, Figure 3.e illustrates a GNSS data stream multiplexed with other data that is written to a single file. This non-GNSS SDR data may be from additional sensors (as shown), and may be written in a proprietary format that may or may not be known.

The specification of metadata parameters for non-GNSS data is outside the scope of the present standardization effort. However, the standard must support adequate information to skip over such non-GNSS data bytes. Since the metadata schema is extensible, it can cover the description of non-GNSS SDR data as needed by specific users.

It is important to note that although we use the term “GNSS SDR data” in this article to generally refer to the type of sampled data for which metadata parameters are defined in the standard, the samples need not correspond to GNSS frequency bands. For example, frequency bands containing RF signals of opportunity are supported as long as they can be represented by the standard’s metadata parameters.

Due to the typically high data rate of GNSS SDRs, some DCSs write data as temporally split files, as illustrated in Figure 3.f. This allows for efficient data management through multiple smaller files compared to a single large file. The metadata for each file must associate the previous and next files in order to represent the sequence. Note that the DCS block shown in Figure 3.f could represent any of those described in topologies a, c, d, and e of Figure 3.

Figure 3.g illustrates spatial splitting of files. Here, the binary data lanes from two or more DCSs are written to separate files. These files may be written by different computer systems. In this case, a non-zero inter-system timing offset, Δt, may exist and must be supported in the standard. Since Δt may or may not be known at time of collection, it represents an example metadata parameter that may be back-annotated into the metadata file after an initial pass of SDR data processing.

Multi-lane DCSs that write each lane to a separate file may also implement temporal file splitting according to 3.f. We call this “spatial-temporal splitting”, and is illustrated in Figure 3.h.

Since multi-stream-multi-file data collection topologies (i.e., (g) and (h) in Figure 3) produce multiple data files applicable to a given time interval, the SDR processor must be “introduced” to the full or partial set of lanes associated with the topology. This is covered by lane selection parameters in the standard.

Metadata Parameters
The metadata standard enables a user to specify a wide range of properties of the binary data file. Some of these properties relate to the data collection scenario itself, such as the time and location, and the type of scenario. Other properties of the dataset relate to the particular RF data that is captured as IF samples, including center frequency and bandwidth. Yet other details such as the IF digitization configuration and data packing can be specified. These are briefly summarized below. Some of the parameters are free-form text, others are integer or floating point variables, and others are enumerations. The primary classes are shown in Figure 4.

The metadata “session” class holds details such as the time, the position, the measurement campaign, and the type of scenario. The “file” class contains a path to the corresponding binary data file, time-stamp info, and a reference to the contained “lane” (described below). A “system” class contains details of the data recording equipment, such as the equipment name, the reference clock frequency, and the antenna details/specifications. The metadata standard allows for the specification of a variety of RF bands, each “band” class includes details of the band’s center frequency, intermediate frequency, bandwidth and group-delay bias.

The actual format of the binary data is detailed in the “lane” class, which allows for the specification of a hierarchy of binary data packing patterns, entitled “blocks”, “chunks” and “lumps”. These classes specify the arrangement of the packed IF data and optional additional data such as sensor measurements or configuration information, and further specify the arrangement of data relative to the standard word-sizes (arrays of byte, short integers, integers, long integers, etc.). The arrangement of the raw IF samples within these blocks is specified by the “stream” class. This class specifies the associated RF band, the sample rate, the quantization, the sample format (real/complex), the encoding of the binary data and the arrangement/alignment of the samples. Together these classes contain sufficient information to unambiguously interpret the packed binary data.

Normative Reference Software
The primary responsibilities of the Normative Reference Implementation Subcommittee are to reduce to practice the conceptual design developed through consensus by the WG into an appropriate set of schema (according to industry best practices) and to develop a compliant software library implementation. The WG was fortunate to receive voluntary participation for this task from individuals representing established commercial GNSS SDR vendors.

The WG decided to work on a publicly available reference software library to be released with the standard. The goal of this effort is to promote early and widespread adoption of the standard by making it easy for vendors and researchers to integrate standard-compliance by integrating the library into their existing software. The scope of this effort is twofold, to develop a metadata interpreter, and to develop a binary data converter using this metadata interpreter.

The first contribution was a library that would produce standard-compliant metadata files from an appropriately-populated data structure, a library capable of reading the content of such files back to a data structure. The second contribution was in the form of a library capable of parsing and converting binary IF data based on an associated metadata description. When combined, it is hoped that these two libraries offer sufficient functionality to enable adoption of the metadata standard in SDRs, or to serve as a benchmark against which implementations of the standard can be validated.

The software is written in C++ and managed using CMake. It has been divided into two libraries, “apilib” implementing the Metadata Interpreter and “converterlib” implementing the binary data conversion, and is accompanied by a selection of utilities, including a data-converter application and a simple test application. The software is available on the Institute of Navigation’s GitHub account here.

The Metadata Interpreter library includes a reader functionality that can parse a metadata file and populate a corresponding metadata object that can then be queried through a selection of member-function calls. Similarly, a metadata object can be instantiated and configured through a selection of member functions and, subsequently, it can be instructed to write a corresponding metadata file.

The converter library can be used for parsing binary data files and interpreting the data according to a metadata file. The basic converter can be adapted to support processing of the converted data streams, and two such adaptations have been implanted in the reference software. The first functionality is depicted in Figure 5, where the data-converter is embedded in a file-converter. The file-converter configures the embedded data-converter using a Metadata Interpreter object, and is capable of parsing a packed binary input file and producing one file per IF data stream in a user-specified data type (int8, int16, float, double, etc.).

A second functionality has been implemented by embedding the data converter in a “front-end”, as depicted in Figure 6. This front-end offers a means of loading short portions of the binary data file and converting them to a user-specified data type (int8, int16, float, double, etc.) while handling details such as sample alignment, when different streams are sampled at different rates.

The software suite includes a selection of example binary datasets and associated metadata files along with a simple MATLAB/Octave script to test the build against reference datasets. To date five different file formats have been included in the repository including a wide range of front-end configurations and data packing variations.

A number of working group members volunteered to perform “blind testing” of SDR data files against draft specifications. This involves exchanging SDR data files and associated metadata specifications among parties and verifying that the files can be fully decoded without additional information. Working group members participating in this activity will comprise the Compliance Verification Subcommittee.

SDR Data Repository
As with most standards and programming projects, they are best understood through good examples. Thus, this page has been created and contains several binary sample files together with metadata files. All file sets have been tested to follow the standard and to be readable by the normative reference software. The binary files typically contain samples over a duration of more than 60 seconds and position fixes have been obtained by at least one software receiver implementation.

Further examples will be added, not only emphasizing new data recording systems but also illustrating different SDR applications such as antenna arrays, reflectometry or ionospheric scintillation analysis. Further, data from GNSS seen only at certain locations of the Earth (e.g., Quasi-Zenith Satellite System [QZSS]) are being considered.

Organizations are encouraged to contact the working group if they have files with formats that are currently not adequately represented in the repository or represent use cases so far not considered. The respective point of contact is given on the web page. The working group will then take care of creating a metadata file (if not yet available), check the correctness and organize the upload.

Additional Resources
[1] Cooklev, T., R. Normoyle, and D. Clendenen, “The Vita49 Analog RF-Digital Interface,” IEEE Circuits and Systems Magazine, Q4 2012.
[2] Favenza, A., and N. Linty, “Exploiting Standardized Metadata for GNSS SDR Remote Processing: a Case Study”, ION GNSS+ 2016, Portland, Oregon (USA), 09/2016.
[3] Lucas-Sabola, V., G. Seco-Granados, J. A. López-Salcedo, J. A. García-Molina, M. Crisci, “Cloud GNSS receivers: New advanced applications made possible”, Proc. Int. Conf. Localization GNSS (ICL-GNSS), pp. 1-6, 2016.
[4] Lucas-Sabola, V., Seco-Granados, G., López-Salcedo, J.A., García-Molina, J.A., Crisci, M., “Demonstration of Cloud GNSS Signal Processing,” Proceedings of the 29th International Technical Meeting of The Satellite Division of the Institute of Navigation (ION GNSS+ 2016), Portland, Oregon, September 2016, pp. 34-43.

The post GNSS SDR Metadata Standard appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

There are many good reasons for getting excited about highly automated vehicles, or HAVs, which is the acronym used by the National Highway Traffic Safety Administration (NHTSA). HAVs can make driving more fuel- and time-efficient. They can significantly reduce traffic congestion and emissions by driving a precise speed, minimizing lane changes, and maintaining an exact distance to neighboring cars. They can also increase accessibility and mobility for disabled and elderly persons.

There are many good reasons for getting excited about highly automated vehicles, or HAVs, which is the acronym used by the National Highway Traffic Safety Administration (NHTSA). HAVs can make driving more fuel- and time-efficient. They can significantly reduce traffic congestion and emissions by driving a precise speed, minimizing lane changes, and maintaining an exact distance to neighboring cars. They can also increase accessibility and mobility for disabled and elderly persons.

Sharing an HAV instead of owning is projected to dramatically reduce a household’s yearly transportation budget, which currently ranges between approximately $8,000 and $11,000 per car. HAVs carry promises not only in improved road mobility, and accessibility, but also in producing architectural and societal changes that can make mass parking spaces and personal car ownership obsolete in urban areas. Above all, HAVs can help improve road safety by preventing car accidents that cause more than 30,000 deaths/year in the United States alone, cost approximately $230 billion/year in medical and work loss costs, and are caused by humans 90% of the time.

Press articles in the 1950s and 1960s predicted that autonomous cars and “electronic highways” would become widely available by 1975. Major milestones in the use of new sensor, computation, and communication technology have recently reenergized the eagerness for HAVs. This first started with the 2005 “DARPA Grand Challenge”, where four different HAVs designed by teams of engineers from industry and academia completed a 132-mile trip across the Mohave desert in less than 7.5 hours with no human intervention. The 2007 DARPA “Urban Challenge” saw six teams autonomously complete a 60-mile course in an urban environment, while following traffic laws. Most teams used a combination of LiDAR, cameras, differential GPS, and computation power that is multiple orders of magnitude higher than what is typically needed for a commercial passenger vehicle. In 2009, Google (now Waymo) began designing and testing “self-driving” cars, which have since accumulated more than three million miles in autonomous mode.

Currently, most car manufacturers have HAV prototype systems and Google, Uber, NuTonomy have HAV pilot testing programs, including fully autonomous systems for public transportation, which, for now, are confined to segregated lanes and geo-fenced areas. Multiple Tier-2 supplier companies have emerged, which specialize in autonomous car technology. In early 2017, 36 companies were registered to test prototype HAV systems on public roads in the state of California.

However, in Figure 1 (for all figures, see inset photo, above right), Gartner’s “2016 Hype Cycle for Emerging Technologies” shows that HAV technology might be at the “peak of inflated expectations”, approaching the “trough of disillusionment”. Hype cycle curves are non-scientific tools that have been empirically verified for multiple example technologies over many years. Two example emerging technologies, commercial unmanned aircraft systems (UAS) and virtual reality, are included in Figure 1 for illustration purposes. The curve’s time scale may differ for each technology. One of many indicators of decreasing expectations on HAVs include a reduction in press coverage and the emergence of first negative news stories, in particular following the May 2016 crash of a Tesla Model S whose autopilot failed to distinguish a white trailer truck from the bright Florida sky. The Model S ran under the trailer causing its roof to be torn off and the operator to lose its life. The car kept going full speed on the side of the road through two fences until it hit a pole and came to a stop.

In parallel, until the end of 2016, Google was providing detailed reports of their self-driving car performance, which were designed to operate in real-world urban environments. These reports contain records of millions of miles driven autonomously, but also acknowledge “disengagements”, i.e., where the operator needed to take over control to avoid collisions. The data shows that HAVs are much more likely to be involved in collisions, even though these collisions are often of lower severity than in conventional human driving [HAVs typically get rear-ended because of their unusual road behavior] (see B. Schoettle, and M. Sivak, “A Preliminary Analysis of Real-World Crashes Involving Self-Driving Vehicles,” Additional Resources). Also, Uber’s autonomous taxis in Pittsburg have a reported rate of one disengagement per mile autonomously driven.

Moreover, the first fielded autonomous systems have revealed new safety threats. In particular, the technology’s functionality, as perceived by the human operator, does not always match the intended operational domain: for example, there have been cases of highway autopilots being used in urban areas and passing red lights without slowing down. In addition, human-machine interaction is at the heart of role confusion (is the operator or the HAV in charge?) of mode confusion (is the HAV in autonomous or manual mode?) and of the operator’s trust in this multimodal system. Misinterpretation may grow even wilder because a given functionality will not achieve the same level of performance across models and manufacturers, and operators may not be aware of the systems’ independently verified safety ratings. And, within the next few years, operators will be expected to anticipate hazardous situations and take over control. Thus, operating an HAV may require more education and different training than driving a car manually.

Current Safety Assessment Efforts
To focus this article, first consider the Society of Automotive Engineer (SAE) International’s classification of driving autonomy levels in Table 1 (see inset photo, above right). Under Levels 0 to 2, the human driver is responsible at all times, either for driving by himself, or for supervising the HAV in autonomous mode and taking control if needed. Under Levels 3 to 5, the system is self-monitoring and the driver is expected to take control, but only if requested by the system. Levels 0-4 provide partial automation under predefined driving modes and circumstances, whereas Level 5 is full autonomy.

The most advanced private car systems are currently Level 2, and pilot programs aim at achieving Level 3, although the mere presence of a kill-switch would imply that the system is actually Level 2. The transition from Level 2 to 3 is a remarkable leap that has significant implications on trust and comfort of human-machine interactions, on legal responsibility allocation between system and driver, and on technical challenges to overcome to guarantee passenger safety.

Over the past four years, the most publicized approaches to demonstrate Level 2 HAV safety have been experimental testing campaigns by Google, Tesla and Uber. Google’s approach to have HAVs drive millions of miles with minimal human intervention has been documented up until 2015. At this time, Google cars have autonomously travelled an impressive three million miles. Tesla’s autopilot is reported to have driven more than 130 million miles – on highways only – before it caused a fatality in May 2016.

In parallel, NHTSA reports about 3,000 billion miles travelled each year on U.S. highways by human drivers, with 30,000 deaths caused by traffic accidents; this corresponds to about one fatality in traffic accidents per 100 million miles driven in the U.S. But, this number accounts for incidents on all roads, in all weather conditions, and for all vehicle ages and types. Thus, a purely experimental, complete proof that HAVs match the level of safety of human driving would take about 400 years at Google’s current testing rate (of approximately 250,000 test miles per year), and would still take many decades if the testing rate increased exponentially. This is assuming that no fatalities occur during that time, that no major HAV upgrade is performed, and that the testing environment is representative of all U.S. roads. Thus, while an experimental proof is conclusive, it is not practical. Other, analytical, methods must be employed to ensure HAV safety.

Research Challenges In HAV Navigation Safety
Multiple technical aspects developed over decades for automated flying could serve as starting points for automated driving systems. Figure 2 shows research areas with overlap between aircraft (in blue) and car (in yellow) applications. Figure 2 is not intended to give a comprehensive list of all aspects of automation, but instead, it shows example technical areas that can be addressed using similar methods in aviation and automotive applications (in the green area). For example:

• performance standards set for software, communication, and electronic equipment are already being compared for aircraft versus cars in the NHTSA report by Q. D. Van Eikema Hommes, Additional Resources.
• the design of aircraft cockpit has been continuously improved over the past few decades, especially for highly-automated Unmanned Air Systems (UAS) with a remote pilot “in-the-box”; few car manufacturers envision futuristic car interiors where humans do not participate in driving, but as long as human-machine interactions are needed, lessons learned in cockpit design to avoid information overload are key. 
• while Automatic Dependent Surveillance-Broadcast (ADS-B) will be mandatory on all aircraft by 2020, a petition for proposed rule making has been issued to mandate Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) by the same date. (ADS-B is a situational awareness system for collision avoidance, through which aircraft share their positions with Air Traffic Control and with other aircraft.) 
• GNSS/INS navigation systems, which are extensively used in safety-critical aircraft navigation, are also being investigated for HAVs.
• overall safety standards also have similarities for aircraft and HAVs, which are discussed again below.

The focus of this article is on navigation safety. In aviation navigation, safety is assessed in terms of integrity (as well as accuracy, continuity, and availability, which are not discussed for brevity). Integrity is a measure of trust in sensor information: integrity risk is the probability of undetected sensor errors causing unacceptably large positioning uncertainty (See RTCA Special Committee 159, “Minimum Aviation System Performance Standards for the Local Area Augmentation System (LAAS), Additional Resources”). This top-level quantifiable performance metric is sensor- and platform-independent, and can thus be used to set certifiable requirements on individual system components to achieve and prove an overall level of safety.

The multiple separate efforts towards achieving Levels 3-to-5 HAVs reveal a compelling lack of coordination towards a common, uniform, quantifiable safety goal. Integrity can be used as an objective performance metric for open, transparent comparison and categorization across manufacturers. It can also provide a governmental regulating agency performance and testing standards for HAV certification, which would help accelerate the development, growth, and maturation of such HAVs, as displayed in Figure 3.

Moreover, the Federal Aviation Administration (FAA) has developed analytical methods to evaluate integrity. This provides the means to:

• quantify safety of existing multi-sensor systems under a variety of operating environments, thereby reducing the need for experimental testing
• allocate safety requirements to individual system components to achieve an overall target level of safety, thereby enabling design for safety 
• perform risk prediction, which is a key operational feature to enable hazard avoidance maneuvers

Several methods have been established to predict the integrity risk in GNSS-based aviation applications, which are instrumental in ensuring the safety of pilots and crew. As an example, Figure 4 illustrates a simplified definition of the integrity risk for aircraft landing applications. The aircraft positioning prediction is uncertain because of sensor measurement noise. An alert limit (AL) requirement box is represented around the predicted aircraft position. This AL is set by the certification authority, i.e., by the FAA in this application. Simply put, the risk of the actual aircraft position being outside the AL box is the integrity risk. (In practice, the most challenging part of risk prediction is to account for potentially undetected sensor faults, such as excessive GNSS satellite clock drift.)

Unfortunately, the same methods do not directly apply to HAVs, because ground vehicles operate under sky-obstructed areas where GNSS signals can be altered or blocked by buildings and trees. In general, the HAV environment is much more unpredictable than the aircraft’s, for reasons that include:

• a changing environment: traffic lights, construction, impact of rain on road adherence, sensor masking and occlusions,
• environmental diversity: intersection topography, road conditions, markings on ground, various traffic signs 
• road users that may interfere with HAV motion: other cars, trucks, pedestrians, bicyclists, etc. 
• comparatively large number of car manufacturers, equipment suppliers, and vehicle models, as well as with shorter model cycles than aircraft, causing wide variations in vehicle age and maintenance levels 
• non-uniform vehicle and road regulations at both the state and federal levels in the U.S. coupled with different international standardization processes.

Thus, HAVs require sensors in addition to GNSS, including laser scanners, radars, cameras, and odometers.

The parallel between aircraft and car applications in Figure 4 illustrates the significant challenge that lies ahead when bringing aviation safety standards to HAVs. It took decades of research and considerable resources to bring the alert limit requirement box down to 10 meters above and below the aircraft using the FAA’s GPS augmentation systems (the Wide-Area Augmentation System and the Local Area Augmentation System). For a car to stay in its lane, the alert limit requirement box must be an order of magnitude smaller, and has to maintain this level of safety in a more dynamic and unpredictable environment.

HAV Taxonomy
Creating a path to successful automated navigation requires an overall methodology to prioritize on imminently achievable objectives, and then expand to more challenging missions. First in this HAV taxonomy, a classification using six SAE autonomy levels has been presented in Table 1. This classification is further refined by segmenting a car’s trip into basic driving competencies, and by specifying the conditions under which a given HAV shall achieve these competencies. A similar classification was made in the early days of GPS-based commercial aircraft navigation safety analysis, where distinctions were made between different phases of flight, weather conditions, vehicle equipment, and airport infrastructure capabilities.

For example, in the early 1990’s, 40% of aircraft accidents were occurring during final approach and landing, and 26% during take-off and initial climb, which only represented an average of 4% and 2% of flight time, respectively. The FAA therefore concentrated their efforts on improving safety during these phases of flight. GPS augmentation systems were designed, with varying capabilities depending on airborne equipment and airport infrastructure, to guide the aircraft under the cloud ceiling, or to bring it all the way to touch-down. Similarly, the “first and last mile” are identified as the most challenging parts of HAV operations, whereas highway auto-drive systems have already been developed and implemented. In its 2016 Federal Automated Vehicles Policy, NHTSA identifies 28 HAV behavioral competencies, which are particularly challenging to meet in the first and last miles of a typical trip. These competencies are basic abilities that an HAV must have to complete nominal driving tasks; they include, for example, lane keeping, obeying traffic laws, and responding to other road users.

To better describe an HAV’s ability, the Federal Automated Vehicles Policy further specifies that basic driving competencies should be available under an HAV’s predefined Operational Design Domain (ODD), described by its geographical location, road type and condition, weather and lighting condition, vehicle speed, etc. The ODD captures the circumstances under which an HAV is supposed to operate safely.

Such classification is key to safety analysis. It can allow HAVs at different stages of their development to be simultaneously fielded, and for them to evolve by expanding their ODDs. The classification can also help in identifying geographical areas where improved road infrastructure is needed for automated operation, similar to airports requiring equipment for instrument navigation to deal with higher traffic density.

Furthermore, standards for electronic equipment, measured by Automotive Safety Integrity Levels, have been issued and compared with the aviation’s Design Assurance Levels (DAL). And, overall system safety levels have been codified, which in aviation account for both the severity and probability of occurrence of an incident, and in automotive applications account, in addition, for “controllability”, which is a measure of how likely an average driver is to maneuver out of a given imminent danger.

All of the above elements: (a) HAV autonomy level, (b) basic driving competency, (c) operation design domain, (d) vehicle electronic equipment, and (e) overall safety risk requirement must be specified to carry out a formal HAV safety analysis. Still missing from the HAV documents are clear guidelines, or example methods, on how to implement these safety requirements.

A Path Towards HAV Navigation Safety
When quantifying the safety of HAV navigation systems, such as in the example displayed in Figure 5, every component of the system including raw sensors, estimator and integrity monitor, and safety predictor, can potentially introduce risk. Unlike aircraft, HAVs require multiple and varied sensors to compensate for GPS signal blockages caused by buildings and trees. These sensor types must be integrated, and new methods to evaluate the integrity of multi-sensor systems must be developed. Furthermore, HAVs must have the ability to continuously predict integrity in a dynamic HAV environment.

In general, research on analytical evaluation of HAV navigation safety is sparse. For example, J. Lee et alia, Additional Resources use the concept of a “safe driving envelope,” but the approach focuses mostly on collision avoidance. The paper by O. Le Marchand, et alia, evaluates ground vehicle navigation, but shows an “approximate radial-error” of tens of meters, far exceeding the necessary sub-meter alert limit. A multi-sensor augmented-GPS/IMU system is used in the paper by R. Toledo-Moreo, et alia with “horizontal trust levels” of 7 meters to 10 meters, still an order-of-magnitude higher than the required HAV alert limit.

Multi-sensor integrity is addressed by M. Brenner, Additional Resources, but for a sensor combination specific to aviation and insufficient for terrestrial mobile robots. Other approaches to multi-sensor integration show promise, but do not provide rigorous proof of integrity. In fact, most publications use pose estimation error covariance as a measure of performance, which is understood as not being sufficient, but is the only metric currently available. Most critically, the metric does not account for fault modes introduced by feature extraction and data association, two algorithms commonly used in mobile robot localization (and discussed again below).

Unlike GPS, which gives absolute position fixes, IMUs, LiDAR, radar, and cameras provide relative displacements with respect to a previous time-step, or with respect to a map. Thus, measurement time-filtering is required, which makes integrity risk evaluation more challenging since past-time sensor errors and undetected faults can now impact current-time safety.

Example LiDAR Navigation Safety Evaluation
While safety quantification for GNSS and GNSS/INS has been rigorously performed for aviation applications, and is being researched for HAVs, navigation safety for LiDAR, radar, camera, and multi-sensor navigation is a widely unexplored research area. To provide a specific example on the research work that lies ahead, we have started developing safety risk evaluation methods for LiDARs. We selected LiDARs because of their prevalence in HAVs, of their market availability, and because of our prior experience. However, the techniques we are developing are general enough that radar, cameras, or any future sensor that returns range data can be substituted.

Raw range data must be processed before it can be used for navigation. One technique, visual odometry, establishes correlations between successive scans to estimate sensor changes in pose (i.e., position and orientation). These processes are highly computationally intensive, and have the same problems as other dead-reckoning techniques, such as wheel odometry over time. Thus, they can become inaccurate or cumbersome for HAVs moving over multiple time epochs. Although proprietary information regarding the use of visual odometry by HAV manufacturers is unavailable, the research literature suggests that it is only used for short time scale operations. A second class of algorithms provides sensor localization by extracting static features from the raw sensor data and associating those features to a map. This is typically done in two steps, as illustrated in Figure 6: feature extraction (FE) and data association (DA). The resulting information can then be iteratively processed using sequential estimators (e.g., Extended Kalman filter or EKF), which has been readily used in many practical applications.

There are several problems that the FE and DA algorithms are addressing. First, landmarks in the environment are unidentified, and their observations are not tagged in a manner similar to a GNSS satellite signal’s Pseudo Random Noise (PRN) number. Thus, the feature extraction algorithm must isolate the few most consistently identifiable, viewpoint-invariant landmarks in the raw sensor data. These features must be identifiable over repeated observations and distinguishable from one landmark to another. Features that are difficult to distinguish from each other can be found easily, but the possibility that the association is incorrect will greatly negatively impact the integrity risk.

Second, range data based on extracted features must match those features with those from a feature database or map. Data association algorithms accomplish this; however, incorrect associations commonly occur. These can lead to large navigation errors, as illustrated in Figure 6, thereby representing a threat to navigation integrity.

FE and DA can be challenging in the presence of sensor uncertainty. This is why many sophisticated algorithms have been devised. But, how can we prove whether these FE and DA methods are safe for life-critical HAV navigation applications, and under what circumstances? These research questions are currently unanswered. The most relevant publications on DA risk are found in literature on multi-target tracking. For example, in the paper Y. Bar-Shalom and T. E. Fortmann, an innovation-based nearest-neighbor DA criterion is introduced, which serves as basis in many practical implementations. The article by Y. Bar-Shalom, et alia, “The Probabilistic Data Association Filter,” provides a detailed derivation of the probability of correct association given measurements. However, this Bayesian approach is not well suited for safety-critical applications due to the lack of risk prediction capability, and to the problem of bounding the a-posteriori probability of association (a similar issue is encountered in the paper by F.C. Chan, et alia. Another insightful approach is followed in the paper by J. Areta, et alia). However, it makes approximations that do not necessarily upper-bound risks, hence do not guarantee safe operation, and it presents exact solutions that can only be evaluated using computationally expensive numerical methods, not adequate for real-time navigation. Also, the risk of FE is not addressed.

In response, we have been developing a new, computationally-efficient integrity risk prediction method to ensure safety of localization using LiDAR-based FE and DA. We have derived a multiple-hypothesis innovation-based DA method that provides the means to predict the probability of incorrect associations considering all potential landmark permutations. (For more details on these methods, see the following four papers in Additional Resources, Nos. 31, 49, 50 and 51.) We also determined a probabilistic lower bound on the minimum feature separation, which is guaranteed at FE, with pre-defined integrity risk allocation. The separation bound can be incorporated in an overall integrity risk equation. This new method was analyzed and tested to quantify the impact of incorrect associations on integrity risk. It showed that the positioning error covariance can be a misleading safety performance metric since cases were found where the contributions of incorrect associations to integrity risk far surpassed that of nominal errors accounted for in the positioning error covariance. In addition, the following key safety-tradeoff was illustrated: the more measurements are extracted, the lower the integrity risk contribution is under the correct association hypothesis, but the higher the other integrity risk contributions become because the risk of incorrect associations increases in the presence of cluttered, poorly-distinguishable landmarks. Finally, being surrounded by many landmarks increases the probability of continuous, uninterrupted navigation. The next step of this research aims at dealing with unmapped and non-static obstacles, and at quantifying the continuity risk of FE and DA.

Conclusion
Looking at the emergence of future HAV technology with the prior experience of aircraft navigation safety provides the means to scale up the challenges that lie ahead in the development of fully autonomous (Level 4 and 5) driverless cars. Many parallels can already be drawn between aviation safety requirements and early HAV standards and regulations. Still, the methods to fulfill these standards and regulations have to be established. If analytical methods are pursued, the following tasks need to be accomplished: (1) establish high-integrity raw sensor measurement error and fault models for non-GPS sensors; (2) develop analytical methods to quantify the safety risk of feature extraction and data association algorithms required in LiDAR, radar, and other pre-processing steps in camera-based localization; (3) design multi-sensor pose estimators and integrity monitors to evaluate the impact of undetected sensor faults on safety risk; and (4) derive, analyze, and experimentally implement integrity risk prediction in dynamic environments.

If these challenges are overcome, one will be able to quantify and prove the performance of an HAV’s navigation system — an essential part of safety. Proving navigation system integrity will also help give humans more confidence to trust HAVs, thus further developing the symbiotic relationship between humans and co-robots. Finally, as HAV technology progresses from driver’s aids such as active brake assist to full autonomous driving, this research is relevant now and will remain essential throughout the evolution of HAV technology.

Additional Resources
[1] Abuhashim, T.S., M.F. AbdelHafez, and M.-A. AlJarrah. Building a robust integrity monitoring algorithm for a low cost gps-aided-ins system. International Journal of Control, Automation, and Systems, 8(5):11081122, 2010.
[2] Ackerman , E., “Self-Driving Cars Were Just Around the Corner—in 1960”, IEEE Spectrum, September 2016
[3] Ackerman, E., “After Mastering Singapore’s Streets, NuTonomy’s Robo-taxis Are Poised to Take on New Cities,” IEEE Spectrum, 2016.
[4] Areta, J., Y. Bar-Shalom, and R. Rothrock, “Misassociation Probability in M2TA and T2TA,” J. of Advances in Information Fusion, Vol. 2, No. 2, 2007, pp. 113-127.
[5] Bailey, T., Mobile Robot Localization and Mapping in Extensive Outdoor Environments. PhD thesis, The University of Sydney, 2002.
[6] Bailey, T., and J. Nieto. Scan-slam: Recursive mapping and localization with arbitrary-shaped landmarks. In Workshop at the Institute of Electrical and Electronics Engineers Robotics Science and Systems (IEEE RSS), 2008.
[7] Bakhache, B., A Sequential RAIM Based on the Civil Aviation Requirements. In Proceedings of the 12th International Technical Meeting of the Satellite Division of the Institute of Navigation (ION GPS 1999), pages 1201–1210, 1999.
[8] Basnayake, C., M. Joerger, and J. Aulde, “Safety-Critical Positioning for Automotive Applications”, Inside GNSS Webinar, 2016.
[9] Bar-Shalom, Y., F. Daum, and J. Huang, “The Probabilistic Data Association Filter,” IEEE Control Systems Magazine, 2009, pp. 82-100.
[10] Bar-Shalom, Y., and T. E. Fortmann. Mathematics in Science and Engineering, chapter Tracking and Data Association. Academic Press, 1988.
[11] Bengtsson, O., and A.J. Baerveldt, “Robot localization based on scan-matching-estimating the covariance matrix for the IDC algorithm,” Robotics and Autonomous Systems, Vol. 44, 2003, pp. 29–40.
[12] Bonanni, R., “WAAS – LPV Airport and Aeronautical Surveys”, ANM Airports Conference, 2006.
[13] Bhuiyan, J., “Uber’s autonomous cars drove 20,354 miles and had to be taken over at every mile, according to documents,” available online here, 2016
[14] Blom, H.A.P., and Y. Bar-Shalom. The interacting multiple model algorithm for systems with markovian switching coefficients. IEEE Transactions on Automatic Control, 33(8):780783, 1988.
[15] Brenner, M., Integrated GPS/Inertial Fault Detection Availability. NAVIGATION, Journal of The Institute of Navigation, 43(2):111–130, 1996.
[16] Chan, F.C., M. Joerger, S. Khanafseh, and B. Pervan, “Bayesian Fault-Tolerant Position Estimator and Integrity Risk Bound for GNSS Navigation,” Journal of Navigation of the RIN, available on CJO2014, doi:10.1017/S0373463314000241, 2014.
[17] Chow, E., and A. Willsky. Analytical redundancy and the design of robust failure detection systems. IEEE Transactions on Automatic Control, 29(7):603614, 1984.
[18] Choukroun, D., and J. Speyer. Mode estimation via conditionally linear filtering: Application to gyro failure monitoring. AIAA Journal of Guidance, Control, and Dynamics, 65(2):632644, 2012.
[19] Clot, A., C. Macabiau, I. Nikiforov, and B. Roturier. Sequential RAIM Designed to Detect Combined Step Ramp Pseudo-Range Error. In Proceedings of the 19th International Technical Meeting of the Satellite Division of the Institute of Navigation (ION GNSS 2006), page 26212633, 2006.
[20] ooper, A.J., A Comparison of Data Association Techniques for Simultaneous Localization and Mapping. PhD thesis, Massachusetts Institute of Technology, 2005.
[21] DARPA, “The Six Finishers of the DARPA Urban Challenge,” available online here, 2007.
[22] Defense Advanced Research Projects Agency (DARPA), “Robots conquer DARPA Grand Challenge,” Press Release, U.S. Department of Defense (DoD), 2005.
[23] Department of Transportation (DOT) National Highway Traffic Safety Administration (NHTSA) “Federal Automated Vehicles Policy: Accelerating the Next Revolution In Roadway Safety,” 2016
[24] Diesel, J., and S. Luu. GPS/IRS AIME: Calculation of Thresholds and Protection Radius Using Chi-Square Methods. In Proceedings of the 8th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GPS 1995), page 19591964, 1995.
[25] Dionne, D., Y. Oshman, and D. Shinar. Novel adaptive generalized likelihood ratio detector with application to maneuvering target tracking. AIAA Journal of Guidance, Control, and Dynamics, 29(2):465474, 2006.
[26] Diosi, A., and L. Kleeman, “Laser scan matching in polar coordinates with application to SLAM,” Proc. IEEE/RSJ IROS, 2005.
[27] Dissanayake, G., P. Newman, S. Clark, H. Durrant-Whyte, and M. Csorba. A Solution to the Simultaneous Localization and Map Building (SLAM) Problem. IEEE Transactions on Robotics Automation, 17(3):229–241, 2001.
[28] Dougherty, M., “Caltrans Leadership in Automated Vehicle Research,” Automated Vehicles Symposium 2017 (AVS2017), San Francisco, CA, 2017.
[29] Dragalin, V.P., A.G. Tartakovsky, and V.V. Veeravalli. The interacting multiple model algorithm for systems with markovian switching coefficients. IEEE Transactions on Information Theory, 45(7):24482461, 1999.
[30] Dragalin, V.P., A.G. Tartakovsky, and V.V. Veeravalli. Multihypothesis sequential probability ratio tests. ii. accurate asymptotic expansions for the expected sample size. IEEE Transactions on Information Theory, 46(4):13661383, July 2000.
[31] Duenas-Arana, G., M. Joerger, and M. Spenko, “Minimizing Integrity Risk via Landmark Selection in Mobile Robot Localization,” submitted to IEEE TRA, 2017.
[32] FAA, “System Design and Analysis,” Advisory Circular AC 25.1309-1A, 1988.
[33] FAA, “System Safety Design and Analysis for Part 23 Airplanes”, Advisory Circular AC 23.1309-1E, 2011.
[34] fars.NHTSA.dot.gov, “Fatality analysis reporting system,” Technical report, NHTSA, 2014.
[35] Federal Aviation Administration (FAA), “Automatic Dependent Surveillance-Broadcast Operations,” Advisory Circular AC No: 90-114A, DoT FAA, 2016.
[36] Federal Highway Administration (FHWA), “Vehicle Positioning Trade Study for ITS Applications”, FHWAJPO-12-064, 2012.
[37] Fenton, R. E., and K. W. Olson “The electronic highway” IEEE Spectrum, 1969.
[38] Forsberg, “NovAtel Establishes Advanced Research Partnership with Illinois Institute of Technology and the University of Arizona,” press release, 2016, available here.
[39] Gartner’s “2016 Hype Cycle for Emerging Technologies” available online here.
[40] Gertler, J., A survey of model based failure detection and isolation in complex plants. IEEE Control Systems Magazine, 8(6):3–11, 1988.
[41] Gitlin, J., “Prepare for the part-time self-driving car,” online at arstechnica.com, 2014.
[42] Google, “Google self-driving car testing report on disengagements of autonomous mode”, available online here, December 2015.
[43] Greenblatt, J. B., and S. Saxena, “Autonomous taxis could greatly reduce greenhouse-gas emissions of us light-duty vehicles,” Nature Climate Change, 5:860–863, 2015.
[44] Greiling Keane, A., “U.S. highway deaths decline for a fifth year, longest streak since 1899,” Bloomberg, Published December 08, 2011.
[45] Halsey III, A., and M. Laris, “Blind man sets out alone in Google’s driverless car,” The Washington Post, 2016.
[46] Hewitson, S., and J. Wang. Extended Receiver Autonomous Integrity Monitoring (eRAIM) for GNSS/INS Integration. Journal of Surveying Engineering, 136(1):13–22, 2010.
[47] Hype cycle curves available online here.
[48] International Organization for Standardization (ISO), “Road vehicles – Functional safety”, ISO 26262, 2011.
[49] Joerger, M., “Carrier Phase GPS Augmentation Using Laser Scanners and Using Low Earth Orbiting Satellites,” Ph.D. Dissertation, Illinois Institute of Technology, 2009.
[50] Joerger, M., M. Jamoom, M. Spenko, and B. Pervan, “Integrity of Laser-Based Feature Extraction and Data Association,” Proc. IEEE/ION PLANS 2016, Savannah, GA, 2016, pp. 557-571.
[51] Joerger, M., B. Pervan, “Continuity Risk of Feature Extraction for Laser-Based Navigation,” Proceedings of the 2017 International Technical Meeting of The Institute of Navigation, Monterey, California, January 2017, pp. 839-855.
[52] Joerger, M., and B. Pervan, “Quantifying Safety for Laser-based Navigation,” submitted to IEEE TAES, 2017.
[53] Kalra, N., and S. Paddock, “Driving to safety: How many miles of driving would it take to demonstrate autonomous vehicle reliability?” Technical Report RR-1478-RC, Rand Corporation, 2016.
[54] Kavanaugh-Brown, J., “Where the Research Meets the Road: Automated Highway Passes the Test ”, Government Technology, 1997. available here.
[55] Kelly, R., and J. Davis, “Required Navigation Performance (RNP) for Precision Approach and Landing with GNSS Application,” NAVIGATION, 1994.
[56] Lee, Y.C., “Analysis of Range and Position Comparison Methods as a Means to Provide GPS Integrity in the User Receiver,” Proc. of the 42nd Annual Meeting of The Institute of Navigation, Seattle, WA, 1986.
[57] Lee, J., B. Kim, J. Seo, K. Yi, J. Yoon, and B. Ko. Automated driving control in safe driving envelope based on probabilistic prediction of surrounding vehicle behaviors. Society of Automotive Engineers International Journal of Passenger Cars – Electronic and Electrical Systems, 8(1):207–218, 2015.
[58] Le Marchand, O., Philippe Bonnifait, Javier Ibaez-Guzmn, and David Btaille. Vehicle Localization Integrity Based on Trajectory Monitoring. In IEEE/RSJ International Conference on Intelligent Robots and Systems, pages 3453–3458, 2009.
[59] Leonard, J., and H. Durrant-Whyte. Directed Sonar Sensing for Mobile Robot Navigation. Kluwer Academic Publishers, 1992.
[60] Li, Y., and Olson E.B. A general purpose feature extractor for light detection and ranging data. Sensors, 10(11), 2010.
[61] Lorden, G., Procedures for reacting to a change in distribution. The Annals of Mathematical Statistics, 42(6):18971908, 1971.
[62] Lu, F., and E. Milios, “Globally Consistent Range Scan Alignment for Environment Mapping,” Autonomous Robots 4, 1997, pp. 333-349.
[63] Madhavan, R., H. Durrant-Whyte, and G. Dissanayake. Natural landmark-based autonomous navigation using curvature scale space. In Proceedings of the Institute of Electrical and Electronics Engineers International Conference on Robotics and Automation (IEEE ICRA), 2002.
[64] Malladi, D. P., and J. L. Speyer. A generalized shiryayev sequential probability ratio test for change detection and isolation. IEEE Transactions on Automatic Control, 44(8):1522–1534, 1999.
[65] Maksarov, D., and H. Durrant-Whyte. Mobile Vehicle Navigation in Unknown environments: a Multiple Hypothesis Approach. In IEEE Proceedings on Control Theory Applications, volume 142, pages 385–400, 1995.
[66] National Transport Safety Board (NSTB), “Preliminary Report, Highway HWY16FH018,” Accident Report ID: HWY16FH018, 2016. available online here.
[67] Neville, K., and K. Williams, “Integrating Remotely Piloted Aircraft Systems into the National Airspace System,” Remotely Piloted Aircraft Systems: A Human Systems Integration Perspective, Wiley, 2017.
[68] Nguyen, V., A. Martinelli, N. Tomatis, and R. Siegwart. A comparison of line extraction algorithms using 2d laser rangefinder for indoor mobile robotics. In Proceedings of the Institute of Electrical and Electronics Engineers/Robotics Society of Japan International Conference on Intelligent Robots and Systems (IEEE/RSJ IROS), 2005.
[69] NHTSA “National motor vehicle crash causation survey: Report to congress,” Technical Report DOT HS 811 059, U.S. Department of Transportation, 2008.
[70] NHTSA, “Federal Motor Vehicle Safety Standards; V2V Communications, Notice of Proposed Rulemaking (NPRM),” DoT NHTSA, 49 CFR Part 571, RIN 2127-AL55, 2016, available online here.
[71] NHTSA “Assessment of Safety Standards for Automotive Electronic Control Systems”, DOT HS 812 285, 2016.
[72] Nikiforov, I., New Optimal Approach to Global Positioning System/Differential Global Positioning System Integrity Monitoring. AIAA Journal of Guidance, Control, and Dynamics, page 10231033, 1996.
[73] Nunez, P., R. Vazquez-Martin, J.C. del Toro, and A. Bandera. Feature extraction from laser scan data based on curvature estimation for mobile robotics. In Proceedings of the Institute of Electrical and Electronics Engineers International Conference on Robotics and Automation (IEEE ICRA), 2006.
[74] Othman, N.A., and H. Ahmad. The analysis of covariance matrix for kalman filter based slam with intermittent measurement. In Proceedings of the 2013 International Conference on Systems, Control and Informatics, 2013.
[75] Page, E.S., Continuous inspection schemes. Biometrika, 41(1-2):100–115, 1954.
[76] Parkinson, B.W., and P. Axelrad, “Autonomous GPS Integrity Monitoring Using the Pseudorange Residual,” NAVIGATION, Vol. 35, No. 2, 1988.
[77] Pfister, S.T., K.L. Kriechbaum, S.I. Roumeliotis, and J.W. Burdick. Weighted range sensor matching algorithms for mobile robot displacement estimation. In Proceedings of the Institute of Electrical and Electronics Engineers International Conference on Robotics and Automation (IEEE ICRA), 2002.
[78] Pfister, S.T., S.I. Roumeliotis, and J.W. Burdick. Weighted line fitting algorithms for mobile robot map building and efficient data representation robotics and automation. In Proceedings of the Institute of Electrical and Electronics Engineers International Conference on Robotics and Automation (IEEE ICRA), 2003.
[79] Radio Technical Commission for Aeronautics (RTCA), “Minimum Operating Performance Standards (MOPS) for Universal Access Transceiver (UAT) Automatic Dependent Surveillance – Broadcast (ADS-B),” RTCA, Washington DC, 2009.
[80] RTCA Special Committee 159, “Minimum Aviation System Performance Standards for the Local Area Augmentation System (LAAS),” RTCA/DO-245, 2004.
[81] RTCA Special Committee 159, “Minimum Operational Performance Standards for Global Positioning System/Wide Area Augmentation System Airborne Equipment,” RTCA/DO-229C, 2001.
[82] Reimer, B., “Revisiting the Topic – The Future is Autonomous Driving – But Are “We” on a Near Term Collision Course?” Automated Vehicle Symposium 2017, (AVS2017), San Francisco, CA, 2017.
[83] Röfer, T., “Using Histogram Correlation to Create Consistent Laser Scan Maps,” Proc. IEEE IROS-2002, Lausanne, Switzerland, 2002, pp. 625-630.
[84] Rogowsky, M., “The Truth About Tesla’s Autopilot Is We Don’t Yet Know How Safe It Is”, Forbes, 2016.
[85] SAE International, “Surface Vehicle Recommended Practice: Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles,” SAE Standard J3016, 2016.
[86] Schoettle, B., and M. Sivak, “A Preliminary Analysis of Real-World Crashes Involving Self-Driving Vehicles,” Report No. UMTRI-2015-34, October 2015.
[87] Sobel, M., and A.Wald. A sequential decision procedure for choosing one of three hypotheses concerning the unknown mean of a normal distribution. The Annals of Mathematical Statistics, 20(4):502522, 1949.
[88] Soloviev, A., D. Bates, and F. van Graas. Tight Coupling of Laser Scanner and Inertial Measurements for a Fully Autonomous Relative Navigation Solution. NAVIGATION, Journal of The Institute of Navigation, 54(3):189 – 205, 2007.
[89] Soloviev, A., Multi-Sensor Fusion for Navigation of Autonomous Vehicles. In Proceedings of the 26th International Technical Meeting of The Satellite Division of the Institute of Navigation (ION GNSS+ 2013), pages 3615 – 3632, 2013.
[90] Soloviev, A., C. Yang, M. Veth, and C. Taylor. Assured Vision Aided Inertial Localization. In Proceedings of the 27th International Technical Meeting of The Satellite Division of the Institute of Navigation (ION GNSS+ 2014), pages 2160 – 2173, 2014.
[91] Sukkarieh, S., E.M. Nebot, and H.F. Durrant-Whyte. A high integrity imu/gps navigation loop for autonomous land vehicle applications. IEEE Transactions on Robotics and Automation, 51(3):572578, 1999.
[92] Toledo-Moreo, R., M. A. Zamora-Izquierdo, B. beda Miarro, and A. F. Gmez-Skarmeta. High-Integrity IMMEKF-Based Road Vehicle Navigation With Low-Cost GPS/SBAS/INS. IEEE Transactions on Aerospace and Electronic Systems, 8(3):491–511, 2007.
[93] Tena Ruiz, I., Y. Petillot, D.M. Lane, and C. Salson. Feature extraction and data association for AUV concurrent mapping and localization. In Proceedings of the Institute of Electrical and Electronics Engineers International Conference on Robotics and Automation (IEEE ICRA), 2001.
[94] Thrun, S., W. Burgard, and D. Fox. A probabilistic approach to concurrent mapping and localization for mobile robots. Machine Learning and Autonomous Robots, 31(5):1–25, 1998.
[95] Thrun, S., W. Burgard, and D. Fox. A real-time algorithm for mobile robot mapping with applications to multi-robot and 3d mapping. In Proceedings of the Institute of Electrical and Electronics Engineers International Conference on Robotics and Automation (IEEE ICRA), 2000.
[96] Thrun, S., “Robotic Mapping: A Survey,” Exploring Artificial Intelligence in the New Millenium. Morgan Kaufmann Publishers Inc., 2003.
[97] Thrun, S., “National Highway Traffic Safety Administration (NHTSA),” keynote presentation, ION GNSS 2007, Fort Worth, TX, 2007.
[98] Van Eikema Hommes, Q. D., “Assessment of safety standards for automotive electronic control systems,” NHTSA Report No. DOT HS 812 285, Washington, DC, 2016.
[99] Waymo, “We’ve reached 3 million miles of selfdriving on public roads! That’s 1 million miles in just 7 months,” available online here, 2017.
[100] White, N. A., P.S. Maybeck, and S.L. DeVilbiss. Detection of interference/jamming and spoofing in a dgps-aided inertial system. IEEE Transactions on Aerospace and Electronic Systems, 34(4):12081217, 1998.
[101] Wikipedia , “Automotive Safety Integrity Level,” 2017. available here.
[102] Williams, S.B., G. Dissanayake, and H. Durrant-Whyte. An efficient approach to the simultaneous localization and mapping problem. In Proceedings of the Institute of Electrical and Electronics Engineers International Conference on Robotics and Automation (IEEE ICRA), 2002.
[103] Willsky, A. S., A Survey of Design Methods for Failure Detection in Dynamic Systems. Automatica, 12:601–611, 1976.
[104] Working Group C ARAIM Technical Subgroup, “Milestone 3 Report,” Technical report, EU-US Cooperation on Satellite Navigation, 2015.
[105] Yoshida, J., “Another Tesla Crash, What It Teaches Us,” EE Times, 2016.

The post Towards Navigation Safety for Autonomous Cars appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

Spoofing of Global Navigation Satellite System (GNSS) signals can have deleterious effects on society given the widespread use and dependence of critical infrastructure on GNSS. However, few commercial receivers have significant anti-spoofing (A/S) mechanisms. Even simple interference events such as jamming and meaconing have resulted in erroneous position outputs from shipboard and airborne receivers (see W. Dunkel et alia; S. Pullen and G. Gao; A. Grant et alia; and A. J. Van Dierendonck in Additional Resources). Spoofing tests have shown that deliberate GNSS spoofing could have significant impact on the GNSS receiver and hence GNSS dependent systems (J. S. Warner and R. G. Johnson; D. P. Shepard et alia). While the extent of the impact is still debated, it is clear that a spoofing event would significantly harm some users. So, the debate over the utility of A/S comes down to the likelihood of spoofing events.

It is clear that GNSS spoofing, outside a laboratory or military setting, has occurred. Recently, GNSS spoofing was observed outside the Kremlin (C. Sebastian) and in the Black Sea (see Goward, Additional Resources). Furthermore, the popularity of location-based games such as Pokémon Go has also induced hackers to build and utilize GNSS spoofers (I. Birnbaum). While the spoofer in Birnbaum uses an expensive GNSS signal generator, other professional security groups have put together GNSS spoofers using low cost software defined radios (SDRs), open source software, and some basic GNSS know-how (see L. Huang and Q. Yang). GNSS spoofing capabilities are no longer solely the realm of navigation experts. As time goes by, spoofing capabilities will get better and costs will only decrease.

There are many motivations to spoof. Ordinary citizens may spoof to aid their gaming, to protect their privacy, or to subvert location based charges (e.g., road tolling) or restrictions. A quick search on the Google Play store shows multiple pages of “Fake GPS” applications. The first application, “Fake GPS Location Spoofer Free,” alone has more than 60,000 reviews as of May 2017. This indicates that many people took the time to not only download and use the app but also to comment on its benefits! There is substantial and growing public interest in spoofing location. Coupling these two factors — the availability of GNSS spoofing equipment or know-how and public interest in spoofing — means we should expect more spoofing incidents in the future. And while critical infrastructure may not be the target for most spoofers, it may fall victim as collateral damage.

We developed and examined a GNSS spoofing detection method via direct comparison of acceleration using commercial inertial sensors. The developed concept allows for comparison of the two sensors without coupling GNSS with an inertial measurement unit (IMU). The design allows for a robust, steady state spoof detection capability that can be developed as an add-on to existing receivers. This article focuses on our preliminary development and demonstration of the concept for aviation.

Background: Prior Art & Developed Technique
Prior Art and Goals
Despite not being a current commercial concern, there is significant literature on GNSS spoofing detection (see Additional Resources). Various researchers have proposed and developed numerous anti-spoofing techniques. Antenna-based techniques use signal properties such as direction of arrival and polarization to detect the presence of spoofing. Internal receiver metrics can be examined for signatures found in spoofing attacks. This includes changes in automatic gain control (AGC) and signal power. The network method checks the received signal against known trusted signals. Redundancy techniques check GNSS measurements against redundant internal or external measures.

While there are many A/S techniques, there is no panacea for spoofing. There is currently no one technique that ideally satisfies all needs. There will likely need to be different solutions for different users, applications, and requirements. As each technique is likely only good against a subset of threats, the overall solution may actually employ several, complimentary techniques to cover all desired threats. Regardless, the techniques employed should have certain characteristics. First, they need to be robust meaning that they catch the threats that they were designed for while having very low false alert rates. Second, they need to be reasonable to implement. This means that they do not significantly change existing receiver designs or add to their cost. A/S needs to be effective but also transparent to the user. It cannot inconvenience users through false alerts or additional, costly complexity. This motivates our investigation of the use of simple inertial-based techniques.

Use of inertial sensors to complement and cross check GNSS is not new. Traditional aviation GNSS/inertial cross-checking algorithms for fault detection have previously been adapted to spoof detection (Y. Liu et alia). Tanil et alia investigated the use of inertials with Kalman filtering to perform spoofing detection in the position domain. These techniques, which require comparisons in the pseudorange or position domain, essentially require GNSS to regularly calibrate IMU results. The deep intertwining of GNSS information to transform IMU results to the position domain limit the trustworthiness of the comparison over time. A spoofer may induce a small GNSS error that causes a bias error in the calibration of the acceleration that can then be slowly exploited. Hence, these spoofing detection methods are considered transient detectors as they only have a limited detection window in which the IMU-derived positions can be considered uncontaminated by GNSS spoof induced errors.

Developed Technique
Overcoming the limited spoof detection window means not deeply intertwining GNSS with the IMU-derived results. Position domain comparison requires regular calibration of the MEMS accelerometer and gyroscope measurements by the GNSS and could cause GNSS spoof induced errors to affect IMU results in a manner that cannot be unraveled. Instead we compare the fundamental IMU outputs of acceleration and rotation rate by aligning GNSS and IMU measurement axes. This alignment is accomplished using GNSS information to approximate attitude. For the study, we compared acceleration as measured by the GNSS and IMU and developed test statistics to help decide if spoofing is present. These tests will have to account not just for errors due to the sensors but also for those due to misalignment of the GNSS and IMU coordinate frames. The benefit of the technique developed is that in uncoupling GNSS outputs from the IMU, we provide an unlimited detection window and steady state detection. It also allows the technique to be implemented as an overlay so that it can be an add-on to an existing receiver.

Any spoofing attack without a good estimate of the vehicle acceleration should be detectable. Even a spoofer that can measure the acceleration remotely or relay a measurement of acceleration from an onboard device may be detectable. This is because the spoofer will incur errors and delays that may be detected provided there are high frequency dynamics. However, there are threats that the technique cannot catch. An attacker with accurate and near real-time knowledge of acceleration can slowly drift the measured position from truth as long as they keep the acceleration error within the allowable detection tolerance. Physical security or complimentary detection techniques may handle these threats.

To be effective, the technique requires a high frequency component of acceleration and predictable attitude. The former represents in cryptographic terms, a one-time pad that a spoofer cannot guess a priori. In flight, there can be many sources of unpredictable acceleration — wind, pilot input to thrust, lowering of the landing gear, etc. Others have considered these items for their ability to provide motion that is difficult for a spoofer to predict (C. Tanil et alia (2015a, 2015b)). Because GNSS alone is used to derive attitude, stable or predictable attitude is desired. Single antenna GNSS measurements cannot estimate some attitude parameters such as roll angle without additional information. Without a reasonable sense of the true attitude, the reference frames between the IMU and GNSS may not be well-aligned and a comparison between IMU and GNSS accelerations cannot be made. While the requirement seems demanding, commercial flights desire stable attitude, especially on approach. This makes sense as the aircraft should be reasonably steady for landing. It should not have much roll and the pitch angle should be small as the aircraft tries to maintain a small, constant glide slope (approximately three degrees). Another time where aircraft attitude is reasonably stable is during cruise, i.e., the majority of any flight. Having established a generally stable attitude over the course of a given flight, we now focus on final approach, as it is the most critical phase of flight.

Critical to the utility of the methodology are two key questions. First, are there adequate motions available for spoof detection using a low cost INS? The motion must be semi-random and significant relative to the capability of the sensors and their errors. This will be examined using flight test data. It must be significant enough to rise above the errors and biases induced by our methodology. The second question is whether we can develop a robust, steady-state test metric for spoof detection given that information.

Data Collection & Testing
While theoretically acceleration from GNSS acceleration and microelectromechanical systems (MEMS) inertials should be suitable for aviation and other transportation, real world errors and biases may result in different performance. We conducted a flight test to gather data to validate our theoretical conclusions and examine flight disturbances.

Data collection equipment
Several instruments were used to collect data for evaluating the utility of a low cost accelerometer for spoof detection. (Please see Manufacturers section for information on the various system components). The receiver and flight test vehicle are shown in Figure 1. The receiver is connected to an external aircraft antenna located on the top center of the body. Normally, GNSS carrier derived velocity would be used to calculate velocity and acceleration. However, the equipment set up was fixed for the test and did not collect this measurement. Instead, dual frequency Precise Point Positioning (PPP) at 10 hertz was used as a proxy with only the Global Position System (GPS) constellation being processed. A smartphone provided the MEMS inertial data. Ideally, the inertial should be tied to the same sampling device as the GNSS. However, due to the fixed set up, the inertial portion of the receiver was not utilized.

Flight Test
A flight test was conducted on August 24, 2016 to collect data for the feasibility of concept. The smartphone was placed on the armrest roughly aligned with the aircraft body axis — it was not collocated with the GNSS antenna though it is located at roughly the same place along the aircraft body. The flight test incorporated several segments representative of the key phases of flight. There are straight and level, coordinated banked turns (in a figure eight pattern), and missed approach segments. The flight and its segments, flown over the period of about 3.5 hours, are shown in Figure 2.

Comparison of Flight Acceleration Data
To compare the GNSS and accelerometer measurements, we must align these measurements and account for gravity. Aligning the measurements means rotating the GNSS measurements to the body frame. We first convert the GNSS positions from Earth centered, Earth fixed (ECEF) to the local east north up (ENU) frame using an initial or representative GNSS position. Then positions are differenced and double differenced to get velocity and acceleration in that frame. This information is used for the comparison and to estimate attitude. The velocity vector in the horizontal direction is used to derive the aircraft heading, which is roughly the direction of the aircraft nose or yaw. If the aircraft is relatively level, such as on approach and in level flight, roll and pitch are small (approximately zero) and adjustments are not necessary. If necessary, the velocity vector in the vertical direction can be used to derive the climb angle which approximates the pitch angle with a bias. Roll may also be derived by assuming a coordinated turn. We do not use roll or pitch estimates in the analysis that follows. The estimated angles are used to derive the rotation matrix to transform GNSS ENU axes to aircraft body axes. Gravity must be accounted for as accelerometers measure specific force rather than acceleration. Hence it will measure gravity whereas GNSS will not. We can either add the acceleration due to gravity, g, set nominally at 9.81 meters per second squared (m/s²), to the GNSS up direction or subtract it from the accelerometer z-axis. Both are equivalent and yield the same equation for acceleration difference. These adjustments result in some residual errors — particularly from residual differences between the accelerometer frame and the adjusted GNSS frame. Additionally, the gravity adjustment can also have errors from variations of gravitational force at different locations and altitudes. With the adjustments, we can calculate the acceleration differences between the sensors. This is shown in Equation (1) where is the acceleration from the sensor (accelerometer or GNSS) along the i-axis.

Equation (1) (see inset photo, above right)

Figure 3 shows the comparison of the accelerometer and GNSS PPP derived acceleration on each axis adjusting for heading only. The comparison is conducted with GNSS and IMU acceleration data that has undergone five seconds of exponential averaging. There are periods where the accelerations are well-matched and other periods where they are not. Generally, they match well during level flight and final approach. They do not match well during the turn section or in climb. This is not surprising as these are periods where the small pitch and roll assumptions are not valid. Estimating and accounting for pitch and roll angles results in better alignment and agreement between the accelerations on all axes. Figure 4 shows the acceleration applying roll estimates. Since most turns were reasonably coordinated, the roll estimates are good and their application results in good alignment.

Comparison of Acceleration Data
The initial analysis uses comparisons of the up body axis during approach — up (GNSS) and z-axis (accelerometer). In Figure 5, the estimated vertical acceleration as measured by GNSS and the accelerometer of the first approach is shown. The acceleration is exponentially averaged over five seconds. The only major difference between GNSS and the accelerometer occurs when the aircraft turns (banks) slightly. The two accelerations have a correlation coefficient of about 0.93. Figure 6 shows the vertical acceleration profile of the second approach. Again the GNSS and accelerometer accelerations are well matched with a correlation coefficient of about 0.96. Also note that the acceleration profile is dissimilar from the first approach. This is demonstrated later when the cross-correlation of the accelerometer accelerations between approaches is calculated.

Figure 7 shows the normalized autocorrelation of the IMU acceleration for the first two approaches, again with five second exponential averaging. The figure shows the (1/e) decorrelation times which range from 2.5 to 3.2 seconds for the approaches. Figure 8 shows the cross-correlation of the second approach with the first and third approaches normalized by the maximum autocorrelation of the second approach. The maximum normalized cross-correlation value over all approaches is about 0.55. The results indicate a fast decorrelation period and no significant cross-correlation between approaches. These results affirmatively answer the first question: Aircraft acceleration measured by low cost accelerometer can provide meaningful comparison with GNSS.

We measured the noise on accelerometer and GNSS acceleration using static measurements of vertical acceleration. Without averaging, the accelerometer showed a mean (μ) and standard deviation (σ) of -0.03 and 0.027 m/s², respectively, and the PPP GNSS acceleration was zero mean with a standard deviation of 1.198 m/s². These statistics are used as the basis of our model bounding variance for the statistical spoof detection tests. With five second exponential averaging, the z-axis accelerometer has a mean of –0.03 m/s² and standard deviation of 0.003 m/s². Similarly, PPP up acceleration was zero mean with 0.028 m/s² standard deviation.

Analysis of Detection and False Alerts
The previous section demonstrated two important qualities. First, low cost accelerometers, not coupled to GNSS, are accurate enough to provide corroborative information to the GNSS-derived movement for aircraft approach. Second, aircraft approaches present useful acceleration signatures that can be used like a cryptographic one-time pad to foil spoofing. The next step is to develop a test for spoofing that can provide robust detection with low probability of false alert. Basic, proof-of-concept monitors were developed using just the accelerometer z-axis and standard statistical testing to demonstrate feasibility. The acceleration comparison suggests that using the z-axis on the accelerometer provides the best information. In future development, other axes and/or sensors may be used either independently or in combination.

Two test statistics are examined and standard hypothesis tests are used to develop monitors based on each test statistic. The first statistic uses the difference in acceleration as measured by GNSS and accelerometer. A spoofed GNSS should experience different accelerations than those measured by the accelerometer. The second statistic examines the standard deviation of the acceleration difference (σ_Δa). The σ_Δa should be larger than the nominal value when the accelerations between the two sensors are not well matched. The second test is less sensitive to a relatively constant bias, such as those resulting from axis misalignment.

The first test statistic, z (mean difference), is shown in Equation (2). It examines the mean difference of acceleration (ȳ) normalized by the model standard deviation, σ. It also accounts for the effect of the maximum nominal bias b. The max function used to incorporate the bias since its sign is not known. The statistic should be bounded by a standard normal distribution provided the model standard deviation and bias are representative. Hence, our threshold test is to flag if z > z_thres. For a 10^-9 probability of false alert (P_fa), z_thres is 6.1. The second test statistic, χ², is shown in Equation (3) with n being the number of samples examined, and s² and σ² being the sample and model variances, respectively. For the initial analysis, n = 8 samples are used to generate the sample variance. The statistic is (central) χ² distributed with (n-1) degrees of freedom (dof). Similarly, our threshold test is to flag when χ² > χ²_thres with χ²_thres being 55.87 for 10^-9 and dof equal to 7 (since n = 8). Both statistical tests depend on the model standard deviation, σ, of the acceleration difference. As such, incorrect modeling affects the monitor performance. If σ is too large, then there will be a larger missed detection rate than modeled. Given the steady state nature of the developed spoof detector, this may be acceptable as there are many chances to catch the spoofer. If σ is too small, the false alert rate will be higher than expected. This is the worse outcome of the two possibilities as it may lead users to distrust the system. So it is better to err on the side of slightly too large. For our testing, the exponential average values are used for the test statistics. The model standard deviation, σ, used is 0.06 m/s² which is twice the root sum squared (rss) of the standard deviation of the accelerometer and GNSS acceleration, as found in the static tests. As the exponential average is used, the static exponential average standard deviations are used. This is shown in Equation (4). A test bias, b, of 0.03 m/s² and n = 8 samples are used.

Equations (2), (3) & (4) (see inset photo, above right)

The statistical tests provide the basic building blocks for the spoof detection monitor. There are several considerations that the monitor must address. One important consideration is minimizing false alerts. Each test may get flagged in non-spoofing situations if our assumptions are not well met. For example, unmodeled attitude can cause large differences in the z-axis accelerometer and up GNSS acceleration. Another consideration is that the tests will not flag during every instant where there is spoofing. For example, the first test will not flag if the spoofed acceleration happens to be within the allowable error tolerance of the true acceleration. This can happen purely by chance or if the acceleration does not vary much and so is easy to anticipate. The monitor should be designed to be robust to these issues. A moving observation window is used primarily to reduce false alerts. Initially a five second window is chosen since this is larger than the decorrelation time. Within the window, each test flag must exceed specified thresholds a certain number of times before the monitor issues an alert. The thresholds may differ for different tests and conditions. Figure 9 shows a general architecture for the spoof detection.

Two overall detection monitors based on these tests are implemented. The simple executive monitoring (EM) indicates spoofing if both detectors indicate spoofing by having their moving sums, Σ₁ and Σ₂, respectively, each exceed a threshold value, Σ_thres. A more nuanced EM leverages the strengths of each test. The EM may alert for each of several different conditions. We developed a multi-condition EM that alerts if the simple EM conditions are met or if the χ² test triggered at a higher threshold, Σ_thres,2 only. This allows us to leverage the power of the χ² monitor to detect spoofing even when the mean difference test is oblivious to it. The mean difference test will not flag for acceleration differences that vary by a small shift in time, whereas the χ² test could flag variation changes. These example executive monitors are shown in Figure 10.

To test the spoof detection monitor, both no spoofing (nominal) and simulated spoofing cases are examined. The nominal case tests the probability of false alert. Testing the nominal case is straightforward and is done with the collected data without modification. To test the spoof detection, we do not need to simulate the spoofing signal. We only need to model the effect of the spoofer on the statistical tests – that is, the acceleration resulting from the spoofing signal. The ability to defeat the monitor is determined by the acceleration that the spoofer can predict. An unsophisticated spoofer may have no knowledge of acceleration and hence its best guess is to assume zero acceleration in the vertical. A sophisticated, worst-case spoofer would accurately know the true GNSS acceleration with a small delay and could generate a spoofed GNSS exhibiting any acceleration profile. While the spoofer can produce many different acceleration profiles with delayed knowledge of the true acceleration, repeating back the true acceleration was found to be a good strategy. This is an extreme spoofing scenario as the spoofer only cares to spoof the acceleration profile without regard to the actual spoofed position. An actual attack would be constrained by the need to generate its spoofed positions.

Figure 11 illustrates an example of the accelerations used for evaluation. The figure shows the acceleration as indicated by the accelerometer, nominal PPP GNSS, and the worst case spoofed GNSS as previously discussed for the first approach. The spoofed case shown assumes that the nominal PPP acceleration is known with a two second delay and a spoofed signal is generated with that acceleration (repeat back). Figure 12 and Figure 13 show the acceleration difference (IMU minus GNSS or spoofed GNSS, top) and performance of each monitor (bottom) for the nominal and spoofed cases, respectively. The bottom of those plots show when each test, the mean difference test (black) and standard deviation difference test (red), was triggered over the course of the approach. A zero value indicates no spoofing while a non-zero value (1.5 and 1 for acceleration difference and standard deviation, respectively) indicates a flag by the specified test. In the nominal case, the standard deviation test flags only once while the mean difference test did not flag. In the spoofing case, each test flags many times on the approach though there are some quiet periods where neither tests flag. Figure 14 shows the number of times each test, the mean difference test (black), standard deviation difference test (red), and the sum for both tests (blue), flags over a moving five second (50 sample) window. The top shows the nominal case while the bottom shows the spoofed case. As desired, there is not much happening in the nominal case. Examining the spoofing case, there are many intervals where the tests flag 20-40 times each or 40-80%. However, there are other intervals where there are no flags. Comparing the time periods where there are spoofing flags to the accelerations shown in Figure 11 suggests that the tests are effective during periods with rapid changes in acceleration. No flags occur during reasonably calm acceleration periods. This is not surprising, as the spoofer can easily approximate the actual acceleration in these periods.

Table 1 shows a summary of the results for the simple and for the multicondition EMs from Figure 10 with a threshold, Σ_thres, of 6% or 3 total test flags in a 50 sample window. For the multi-condition EM, the Σ_thres,2 used is 12% or 6 flagged instances. The table shows the percentage of time spoofing is alerted by each EM and time from start to first detection presented for all four approaches and for different cases: nominal, a spoofer with no knowledge (assuming zero acceleration), and the repeat-back spoofing cases. The repeatback spoofing cases are conducted with one-half- and two-second information delay. In the table, any non-zero detection percentage indicates that the EM has generated a spoofing alert during the approach. Hence, the multi-condition EM catches all simulated spoofing cases shown. Additionally, the monitor alerts within about 13 seconds of the start of the approach and spoofing with the exception of Approach 1. This time to first detection (TFD) is a function not just of the monitor but also of the dynamics of the aircraft. With little variation in motion, it is easy for an attacker to predict the acceleration profile and hence remain concealed to the monitor. As seen in Figure 11, Approach 1 does not have much vertical acceleration variation initially. Hence it has high TFD. The simple EM can catch the longer delay (two second) spoofing attack but with a larger TFD. With a shorter delay, the simple EM may not alert throughout the entire approach as the acceleration difference monitor never flags. This is because the acceleration is continuous and does not change rapidly over a short period of time. Thus, with very small delays, difference between the actual and spoofed acceleration can be small and always remains within the tolerances specified by the low probability of false alert. Similarly, the percentage of time the monitor detects spoofing also depends on the dynamics of the flight. For example, the multi-condition EM detects the repeat-back spoofer with half-second delay between 14.2 to 50.3% of the time.

Another important result is that there are no false alerts in any case with the exception of Approach 4 with the multiple condition EM. The cause of the false alert was found to be dropouts in the GNSS measurements, which caused outlier GNSS accelerations for a few seconds. The result of the drop-out, which was exponentially averaged with other measures, can be seen in Figure 15 which shows the accelerations from the accelerometer, GNSS, and spoofer. The standard deviation monitor flagged the resulting jump. Hence, the false alert was due to a data issue rather than the monitor itself. The detection architecture should be designed to manage data handling errors.

Conclusions
The results provide good indication that a low cost IMU can be useful for spoofing detection during critical phases of flight. It demonstrated unique random vertical accelerations experienced on aircraft approach. Furthermore, it found that a good comparison between GNSS and IMU derived acceleration on approach and cruise can be made. Other segments of flight may be used provided we can derive a reasonable attitude estimate without inadvertently allowing a GNSS spoofer to contaminate our IMU results. Approaches having more high frequency and high amplitude accelerations result in better detection. The acceleration differences were used as the basis for a simple and multi-condition executive monitor for spoofing. These EMs demonstrated their spoof detection capabilities and their ability to limit false alerts using collected flight test data. Preliminary results show that monitoring can be designed to detect spoofing on all four approaches tested. Time to first detect depends on both the monitor design and aircraft dynamics. Fast detection (< 10 seconds) can be achieved especially if there are high amplitude and frequency accelerations. Many more flights will be needed to validate the performance results.

The analysis conducted provides only a preliminary feasibility demonstration and there is still much to be done. One area for future work is fault-tolerant design. The detection architecture needs to determine when it is suitable for use – i.e., when the attitude assumptions are valid. While the analysis conducted leverages some special characteristics of flight, other test measurements conducted have shown that this technique may be suitable for other transportation such as railways and automobiles. Both automobile and rail have additional characteristics that can be leveraged.

Acknowledgements
The authors thank FAA Navigation Programs and the Stanford Center for Position Navigation and Time (SCPNT) for supporting this work. We also thank the FAA Technical Center and Stuart Riley of Trimble Navigation for their help.

Disclaimer
The views expressed herein are those of the authors and are not to be construed as official or reflecting the views of the Federal Aviation Administration or Department of Transportation.

Additional Resources
[1] Birnbaum, I.,“‘Pokémon Go’ Players Are Spoofing GPS Locations to Catch’ Em All,” Motherboard, July 8, 2016
[2] Borowski, H., Isoz, O., Eklöf, F. M., Lo, S., and Akos, D., “Detection of False GNSS Signals using AGC,” GPS World, April 2012
[3] Brenner, M., “Integrated GPS/Inertial Fault Detection Availability,” NAVIGATION, Volume: 43, Issue: 2, pp. 111-130, 1996
[4] Chen, Y.-H., Juang, J.-C., Seo, J., De Lorenzo, D., Lo, S., Enge, P., and Akos, D., “Design and Implementation of Real-time Software Radio for GPS/WAAS Controlled Reception Pattern Antenna Array Adaptive Processing,” Sensors, Volume: 12,Issue: 10, pp. 13417-13440, 2012
[5] Chen, Y.-H., Lo, S., Akos, D., De Lorenzo, D., and Enge, P., “Getting Control: Off-the-Shelf Antennas for Controlled-Reception-Pattern Antenna Arrays,” Innovation Column, GPS World, February 2013
[6] Chen, Y.-H., Rothmaier, F., Akos, D., Lo, S., and Enge, P., “Towards a Practical Single Element Null Steering Antenna,” Proceedings of The Institute of Navigation International Technical Meeting, Monterey, CA, January 2017
[7] Diesel, J. and Dunn, G., “GPS/IRS AIME: Certification for Sole Means and Solution to RF Interference,” Proceedings of the 9th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GPS 1996), Kansas City, MO, September 1996
[8] Diesel, J. and King, J., “Integration of Navigation Systems for Fault Detection, Exclusion, and Integrity Determination—without WAAS,” Proceedings of The Institute of Navigation National Technical Meeting, 1995
[9] Dunkel, W., Weber, O., Butsch, F., “GNSS Interference Detection with GIMOS,” 11th International GBAS Working Group Meeting (I-GWG-11), Osaka, Japan, February 24, 2011
[10] Goward, D., “Mass GPS Spoofing Attack in Black Sea?,” The Maritime Executive, July 11, 2017
[11] Grant, A., Williams, P., Ward, N., and Basker, S., “GPS Jamming and the Impact on Maritime Navigation,” Journal of Navigation, Volume: 62, Issue: 02, April 2009
[12] Gross, G. and Humphreys, T. E., “GNSS Spoofing, Jamming, and Multipath Interference Classification using a Maximum-Likelihood Multi-Tap Multipath Estimator,” Proceedings of the International Technical Meeting of The Institute of Navigation, Monterey, CA, January 2017.
[13] Heirich, O., Robertson, P., Garcia, A. C., Strang, T., “Bayesian Train Localization Method Extended By 3D Geometric Railway Track Observations From Inertial Sensors,” 15th International Conference on Information Fusion, July 2012
[14] Huang, L. and Yang, Q., “GPS Spoofing, Low cost GPS Simulator,” DEFCON 23, August 2015
[15] Key, E. L., “Techniques to Counter GPS Spoofing. Internal memorandum,” USA: MITRE Corporation, February 1995.
[16] Khanafseh, S., Roshan, N., Langel, S., Chan, F.-C., Joerger, M., Pervan, B., and “GPS Spoofing Detection Using RAIM with INS Coupling,” Proceedings of Proceedings of The Institute of Navigation/IEEE Position Location and Navigation Symposium (PLANS), Monterey, CA, pp. 1232-1239, May 2014
[17] Last, D., Grant, A., and Ward, N., “Demonstrating the Effects of GPS Jamming on Marine Navigation,” 3rd GNSS Vulnerabilities and Solutions Conference, Baška, Krk Island, Croatia, September 5-8, 2010
[18] Levin, P., De Lorenzo, D. S., Enge, P. K., Lo, S. C., “Authenticating a Signal Based on an Unknown Component Thereof,” US Patent # 7,969,354, June 28, 2011
[19] Liu, Y., Fu, Q., Liu, Z., and Li, S., “GNSS Spoofing Detection Ability of a Loosely Coupled INS/GNSS Integrated Navigation System for Two Integrity Monitoring Methods,” Proceedings of The Institute of Navigation International Technical Meeting (ITM), Monterey, CA, January 2017
[20] Lo, S., Chen, Y. H., Barrows, A., Reid, T., Perkins, A., Jan, S. S., and Enge, P., “Using Traffic Information Services Broadcast (TIS-B) Signals for Aviation Navigation,” Proceedings of The Institute of Navigation/IEEE Position Location and Navigation Symposium (PLANS), Savannah, GA, April 2016
[21] Lo, S., De Lorenzo, D., Enge, P., Akos, D., Bradley, P., “Signal Authentication: A Secure Civil GNSS for Today,” Inside GNSS, September/October 2009
[22] Manickam, S. and O’Keefe, K., “Using Tactical and MEMS Grade INS to Protect Against GNSS Spoofing in Automotive Applications,” Proceedings of the 29th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2016), Portland, OR, pp. 2991-3001, September 2016
[23] McMilin, E., “Single Antenna Null-Steering for GPS & GNSS Aerial Applications,” Ph.D. Dissertation, Stanford University, March 2016
[24] Psiaki, M. L. and Humphreys, T. E., “GNSS Spoofing and Detection,” Proceedings of the IEEE, 2016.
[25] Pullen, S. and Gao, G., “The Impact of Uninformed RF Interference on GBAS and Potential Mitigations,” Proceedings of the ION International Technical Meeting (ITM), Newport Beach, California, January 2012
[26] Sebastian, C., “Getting Lost Near the Kremlin? Russia could be ‘GPS Spoofing’,” CNN Tech, December 2, 2016
[27] Shepard, D. P., Bhatti, J. A., Humphreys, T. E., Fansler, A. A., “Evaluation of Smart Grid and Civilian UAV Vulnerability to GPS Spoofing Attacks,” Proceedings of the 25th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS 2012), Nashville, TN, pp. 3591-3605, September 2012
[28] Swaszek, P. F., Pratz, S. A., Arocho, B. N., Seals, K. C., Hartnett, R. J., “GNSS Spoof Detection Using Shipboard IMU Measurements,” Proceedings of the 27th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2014), Tampa, FL, pp. 745-758, September 2014
[29] Tanil, C.,Khanafseh, S., Joerger, M., and Pervan, B., “Kalman Filter-based INS Monitor to Detect GNSS Spoofers Capable of Attacking Aircraft Position,” Proceedings of The Institute of Navigation/IEEE Position Location and Navigation Symposium (PLANS), Savannah, GA, pp. 1027-1034, April 2016
[30] Tanil, C., Khanafseh, S., and Pervan, B., “Impact of Wind Gusts on Detectability of GPS Spoofing Attacks Using RAIM with INS Coupling,” Proceedings of The Institute of Navigation 2015 Pacific PNT Meeting, Honolulu, HI, pp. 674-686, April 2015 (2015a)
[31] Tanil, C., Khanafseh, S., and Pervan, B., “GNSS Spoofing Attack Detection using Aircraft Autopilot Response to Deceptive Trajectory,” Proceedings of the 28th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2015), Tampa, FL, pp. 3345-3357, September 2015 (2015b)
[32] Tanil, C., Khanafseh, S., and Pervan, B., “An INS Monitor Against GNSS Spoofing Attacks During GBAS and SBAS-assisted Aircraft Landing Approaches,” Proceedings of the 29th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2016), Portland, OR, pp. 2981-2990, September 2016
[33] Thompson, R. J. R., Cetin, E., and Dempster, A. G., “Evaluation of Relative GPS Timing Under Jamming Conditions,” Proceedings of the 25th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS 2012), Nashville, TN, pp. 717-730, September 2012
[34] Van Dierendonck, A. J., “GPS Re-Radiator Issues,” Presentation to the US GPS Industry Council, October 2005
[35] Waid, J. and Fly, B., “Tactical HIGH™ – Solution Separation Methods Applied to the Warfighter Environment,” Proceedings of The Institute of Navigation 60th Annual Meeting, Dayton, OH, 2004
[36] Warner, J. S. and Johnston, R. G., “Think GPS Offers High Security? Think Again!,” Business Contingency Planning Conference, Las Vegas, NV, May 23-27, 2004

The post Consumer Mass Market Accelerometers for GNSS Anti-Spoofing appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

Collision avoidance will be more practically and universally achievable, even in skies crowded with unmanned aerial vehicles (UAVs), if the aviation community takes advantage of the raw measurements already present in today’s GPS receivers, but largely ignored in favor of using GPS position coordinates (see “The Advantages of Raw Measurements” sidebar, below.) Changes to existing aviation equipment could enable aircraft to estimate the flight paths of other aircraft far more accurately enabling safer operations even under impaired conditions. The use of the raw data, once integrated into standardized flight protocols, could dramatically help prevent midair accidents even if one or both of the aircraft is unmanned.

This flight-validated approach uses established algorithms, readily available universal access transceivers (UATs) and existing communication message formats (as described by P. Duan et alia in Additional Resources). There is one essential departure from current practice: including the raw pseudorange and carrier phase data within the automatic dependent surveillance-broadcast (ADS-B) messages in place of the derived position coordinates. This same approach could be applied with great benefit should a separate system, similar to ADS-B, need to be developed to support UAV operations.

For more than 50 years it has been feasible to combine intermittent partial data – of different types at varying accuracies with different sensitivities from different directions at different times – and extract all benefit therein. The seemingly unspectacular step of using raw measurements in the message opens the door to using powerful, widely understood methods for predicting flight paths — and therefore the points of potential collision — and handling situations where position determination is hampered due to jamming or because there are not enough satellites in view. It should be possible to do all of this for aircraft at various altitudes without building and certifying extensive new radar or other ground infrastructure.

This paper primarily involves GPS and airborne operation, but a half-century of experience combining data enables this technique to be dramatically extended. Integration of different sensors (eLoran, DME, etc.) is straightforward; a claim that has been verified and documented.

Raw Measurements Improve Estimates
For a host of reasons, techniques using raw measurements — which are present in any navigation sensor — will outperform by orders of magnitude techniques relying on position reporting. Differential GPS (DGPS) owes its spectacular success to its use of raw measurements. A Kalman tracker uses weights based on an extensive array of data. There are across-axis correlations between error components in different directions, between components of position and velocity, etc. — and the sensitivity of each individual observation to every one of those components is taken into account.

Unfortunately none of those features can be used when starting with coordinates derived from raw measurements. Since ADS-B link bandwidth can’t hold its existing content plus all that correlation information, ADS-B messages contain no correlations — but airborne computation armed with a history of raw measurements can deduce all. The contrast could hardly be more compelling.

The extended squitter message can remain unchanged except for the replacement of position and velocity by raw measurements. Since navigation systems commonly allow multiple message types, however, position reports are not strictly ruled out. Occasionally another message type could be used to broadcast the position for, say, track file initiation. To realize the performance potential, however, messages containing raw measurement data would be far more frequent.

But why go to the trouble of changing an established methodology (that is position reporting) for one based on Kalman filtering considerations? There are multiple reasons. Probably most obvious, instantaneous position is fleeting for anything airborne; data must be combined. Even satellite navigation uses several observations to get a position fix (i.e., four satellites in air). Another reason stems from the nature of the position that must be determined: To support collision avoidance the position information for one object must be determined relative to the other objects in nearby space and projected into the future. Using raw data makes these calculations far, far more accurate.

Consider, for example, a pair of position coordinates, one with perfect longitude but a kilometer of error in the north direction, and the other with exact latitude but its east/west position is off by a kilometer. Averaging them gives “only” 500 meters of error in both!

Even if different tolerances of different position reports are taken into account, ignoring wide variations in sensitivity and correlation parameters is ruinous. There are no tight velocity accuracy requirements specified for ADS-B as a result.

Accurate Velocity Essential
In fact the velocity requirement for ADS-B is loose in multiple ways. For characteristics that matter in regard to collision avoidance, velocity is a vector — a vector relative to other objects in nearby space regardless of those objects’ latitude or longitude — and with errors having statistical properties. Error values of several meters/second, even as much as 10 meters/second, have been published in connection with ADS-B. A more subtle point is that even a substantially lower 1 meter/second error value is dangerous statistically. Without detailed elaboration, this much needs to be recognized: extreme value theory (EVT) shows that, even if all errors were Gaussian, mixed Gaussian probability offers far less assurance than intuition would suggest. Instances of exceeding 10 sigma cannot be discounted (see Farrell J., and F. van Graas, Additional Resources). To ignore that is to accept an excessive and unsafe risk; “unlikely” is often not unlikely enough. A 10 meter/second error is therefore not too farfetched to consider. A sequence of position reports can suffice for transoceanic flight but not within crowded airspace.

Avoiding mid-air crashes requires a fresh look at system priorities. For example it is not critical to have highly accurate position reports. For airliners with wingspans tens of meters long moving at hundreds of kilometers per hour, precise position is fleeting and unnecessary. A few meters of current position error will be insignificant and working to refine that position error would be pointless.

It is essential, however, to have highly accurate velocities. The product of (velocity error) × (time to closest approach) is dominant when it comes to collision avoidance. Instead of meters per second velocity accuracy, airliners need centimeters/second accuracy; otherwise the projected position over time is so crude as to be useless for collision avoidance.

Collision avoidance demands assurance of sufficient distance at time of closest approach. That clearly requires accurate knowledge of velocity — specifically the relative velocity vector. Stitching coordinates together from reports of latitude + longitude + altitude (“LLH”) cannot deliver that, which explains why ADS-B does not promise good velocity. Errors of 10 meters/second have been published with little elaboration (not relative, not vectorial, and with no statistical boundaries). If the closest approach is a minute away, then even the most elementary arithmetic assigns 600 meters of uncertainty to that future position.

Consider, for example, two aircraft flying at the same altitude: “Ownship” at location O with velocity VO and “Another-ship” at location A with velocity VA (see inset photo, above right). They are instantaneously separated by vector R which, for closing scenarios, is shrinking. The closest approach will occur at time T when the component of relative velocity (vector difference VA – VO, not shown) parallel to R passes through zero — the perpendicular component is miss distance. That simple scenario has appeared in countless context-dependent forms (e.g., with intruder at A and evader at O in or with target at A and an interceptor or projectile at O in military operations). Determination of T and minimum separation distance follows easily from relations just stated, readily superseded whenever maneuvers subsequently change either velocity.

Now to confirm the case for precise velocity: rather than current position, collision avoidance requires accurate future position (i.e., at time T). When position is projected T seconds ahead, a 10 meter/second velocity error will cause that predicted future position to be in error by 10 T meters. With that much error in each of two horizontal axes that product will be squared, producing an unacceptably large area of uncertainty. Trying to steer away from an unknown place has no meaning. The Traffic Collision Avoidance System (TCAS) uses climb/dive maneuvers instead. Imagine that becoming a commonplace event as the skies fill with unmanned aircraft.

There are many facets to this subject, but the simple, fixed-altitude case above is useful for establishing some fundamentals:

• Time T used above matches the “tau” of the traffic collision avoidance system (TCAS) only on a collision course
• Effects of current position error matter far less than velocity error 
• Centimeter/second accuracy for velocity rather than meters/ second can enable horizontal evasive strategies 
• Longer values of T (earlier evasive action) is also thereby made possible 
• Earlier evasive action is highly preferable to TCAS’s abrupt violent maneuvering 
• TCAS cannot act early because valid decisions require accurate tracks 
• TCAS tracks are informed by accurate range but very crude cross-range data 
• TCAS cross-range information improves only as the sightline rotates 
• Sightline rotation increases at close range — exactly the waterloo for collision avoidance! 
• The precise satellite navigation data used in the previously mentioned P. Daun et alia article (Additional Resources) provides full 3-D tracks quickly 
• Requisite speed changes and T have been quantified for many cases (see Farrell, J., “Collision avoidance by speed change,” Additional Resources).

Raw Measurements Improve Tracking
As noted in the article on airport surface surveillance (J. Farrell and E. McConkey, Additional Resources) computers now can easily maintain integrated track files for every participant in any scenario. Even in the 1970s two missiles plus two aircraft were simultaneously tracked in real time with an electronically steered radar antenna at White Sands. The estimation algorithms in the White Sands case were fed by raw observations (range, azimuth and elevation in that case) — never with coordinate pseudo-measurements — and tracking from high dynamic platforms with “Ownship” navigation is a straightforward extension of tracking from a stationary location.

Today’s computing capabilities readily enable each participant to maintain a bank of extended Kalman filters (EKFs) with a separate track file for each participant and with every participant having a designated slot in the sequence of transmitted messages from all the participants. The full set of participants should include every object that could be involved in any collision. The track file in any participant’s database is not tied to coordinates; it’s scalar. From those scalars each participant can construct a set of vectors and all those vectors will be correct and can be expressed in his own perceived reference. If that perception differs from the other participants’ (due to misalignments or even a different datum), performance does not suffer one iota.

Air-to-air tracking has always placed the Ownship described in the example above at the center of its “own little world” without any degradation. What matters is relative state (position, velocity…) in Ownship’s “own little world” expressed and maintained consistently the same way for all. Sharing satellite navigation data with others will not introduce any error since those measurements are scalar —unattached to any coordinate frame. If the presence of one participant with overriding authority must be identified, one of the participants could be a tower. With the exception of the tower, if there is one, moving participants would make path adjustments with each message received as the scenario unfolds. Those smaller, repeated adjustments over time will prove far less abrupt than making a start-from-scratch change at close range.

Dealing With Too Few Satellites
Using the raw data also enables the development of track files in situations where there are not enough GPS satellites in view. In fact, in some urban canyon scenarios it is possible to have situations where there are never enough satellites available with good enough geometries.

As things now stand, if an aircraft’s GNSS receiver does not have enough satellites in view it is not able to determine its position and therefore has nothing to broadcast on ADS-B. That is a scandalous waste of very accurate information. Raw data measured every second or so will give you a far better track file than the usage of GPS coordinates. Stitching coordinates together to get velocity gives totally inadequate performance. That is why ADS-B, even with all the ADS-B Out and ADS-B In information, will not provide accurate velocity.

UAV-Specific Considerations
While UAVs will be responsible for taking evasive action, they will be less burdened in other respects. Their lower speed affords multiple advantages: more time for evasion, track file initiation at short range (allowing operation at low power) and the ability to make tighter turns. All of these factors make sense and avoid easier for UAVs than it is for fast-moving airliners.

Nor will UAVs require the sophistication used by P. Duan et alia in Additional Resources, which used 1-second changes in meticulously prepared carrier phase measurements. Many satellite navigation receivers don’t use carrier phase but all have pseudoranges — those will suffice as long as they are made available with appropriate time stamps. Also, decimeters/second rather than centimeter/second velocity error will be acceptable — again because of a UAV’s slower speed. With evasion by acceleration or deceleration, for example, the simple program (see J. Farrell, “Collision avoidance by speed change,” in Additional Resources) can just have different parameters. Finally, evasion strategy won’t be limited to speed changes; descent or turns can be used in some circumstances.

The Challenge
Though the advantages of using raw measurement are clear, change is not easy. Using position, and over recent decades GPS-derived position, in ADS-B messaging has long been the established approach. However the integration of unmanned aircraft is such a monumental challenge that new techniques and air traffic management systems for UAVs are being considered. Incorporating raw measurements not only offers a capability that supports safe UAV integration, but offers real advantages to manned flight operations as well — and there is a rock-solid track record supporting both double differencing (see “Double Differencing” sidebar, below) and all modes of tracking (air-to-air, air-to-surface, surface-to-air, surface-to-surface) by adaptive modern estimation.

Conclusions and Recommendations
Working with the raw measurements instead of relying only on the position calculated using those measurements makes it possible to apply the techniques that made differential GPS so spectacularly successful. This approach also opens the door for the integration of data from information sources completely different from GNSS and from each other. Raw measurements offer the only way to achieve true integration with systems like DME, eLoran and Iridium and, especially for cooperating UAVs, signals-of-opportunity (see R. Kapoor et alia, Additional Resources).The scope can also be extended to include observations of nonparticipants (see Fig. 9.4 in J. Farrell, “ GNSS Aided Navigation and Tracking – Inertially Augmented or Autonomous,” in Additional Resources).

The improvements in situational awareness are dramatic enough to suggest redefining availability and continuity of operation. Less obvious but equally decisive is how this approach strengthens integrity. Every individual measurement can be acceptance-tested — directly, easily, and independently of all others, supported by demonstrated equivalence to rigorous, widely accepted parity methods (see “Integrity Testing: Ultra-simple And Rigorously Validated” sidebar, below).

These dramatic improvements do not require new discoveries or the invention of new equipment. A revision of the ADS-B message content, hopefully via a software update — and the inclusion of raw measurements in any new system developed to support UAVs — will enable a host of spectacular benefits from already readily available data. In fact the use of raw measurements is so promising that SAE International has begun developing standards to bring this approach into the mainstream.

There also are documented non-proprietary navigation algorithms already available that make it possible to tap the value of the raw measurements. These algorithms could help keep costs down and speed the launch of a pilot project to test this approach, especially in the case of unmanned aircraft. There is enormous commercial, political and regulatory pressure to integrate UAVs into the national airspace. A pilot project could support both manned and unmanned aviation by strengthening reliability and robustness while boosting accuracy and integrity — thereby helping keep aircraft out of each other’s way.

An old movie scene showed Bob Hope trudging through a desert, desperately uttering “water, water” — then finding himself waist deep in a stream moments later, mumbling “mirage, mirage.” The advantages of using raw measurements for ADS-B and systems similar to ADS-B are not a mirage. Between what we know and what we do is a wide gulf. Let’s close it.

Appendix—Additional Topics
Two separate but related articles from a recent Institute of Navigation newsletter discuss important developments in GPS/GNSS interfacing. Starting on page 1 and continued on page 7, the first describes major improvements in Android handsets. The second, on pages 14-15, announces formation of a Society of Automotive Engineers (SAE) International working group, which will work on the standards cited in the Conclusions and Recommendations section of this article, ensuring the extension of benefits to the vast majority of devices. (SAE International is a global association of more than 128,000 engineers and related technical experts in the aerospace, automotive and commercial- vehicle industries.) These were preceded by other publications emphasizing the benefits offered by working with measurement data. One, more than 25 years old (J. Farrell and F. van Graas, Additional Resources) was in fact preceded by an obscure (1977) NAECON paper. Two more recent videos here and here plus a presentation offer additional background.

The centimeter/second residuals achieved in flight test described previously by P. Daun et alia, in Additional Resources, were obtained by using sequential changes in carrier phase measurements measured once a second. Unlike the carrier phases themselves, 1-second changes in them are interoperable (i.e., regardless of different timing and/or geoid conventions used for separate constellations) and immune to catastrophic error (see links to https://jameslfarrell.com content in Additional Resources). Furthermore, because two main sources of propagation error change very little over a second, there is no need for a mask angle — a trait that benefits geometric dilution of precision (GDOP) for velocity.

Additional Resources
1. Bayliss, E., R. E. Boisvert, M. L. Burrows, and W. H. Harman, “Aircraft surveillance based on GPS position broadcasts from Mode-S beacon transponders,” ION-GPS94.
2. Duan, P., M.U. De Haag and J. Farrell, “Flight test results of a measurement-based ADS-B system for separation assurance,”, NAVIGATION, Journal of the Institute of Navigation, Volume 60, Number 3, 2013, pp. 221-234; Abstract
3. Farrell, J. and E. D. McConkey “Quantum improvement in airport surface surveillance,” IONNTM, 1998
4. Farrell, J., E. D. McConkey, and C. G. Stephens, "Send measurements, not coordinates," NAVIGATION, Journal of the Institute of Navigation, Volume 60, Number 3, 1999, pp.203-215). 
5. Farrell, J., GNSS Aided Navigation and Tracking — Inertially Augmented or Autonomus, American Literary Press, 2007
6. Farrell, J., "Collision avoidance by speed change," COORDINATES Volume VIII Number 9, Sept. 2012, pp. 8-12
7. Farrell, J., “Letters: Get a Start on GNSS Interoperability Now,”
8. Farrell J. and F. van Graas, “That all-important interface,” James L. Farrell and Frank van Graas, ION-GPS90
9. Farrell J., and M. L. Farrell, “ADSB (2nd-) Best Foot Forward?” Journal of Air Traffic Control, Summer 2008, 44 17-18.
10. Farrell J. and F. van Graas, “Containment Limits for Free-Inertial Coast,” ION-GNSS2010
11. Farrell, J., "Single-Measurement RAIM"
12. Farrell, J., "Send Measurements, Not Coordinates" — pages 14-15
13. Farrell, J., "Dead Reckoning by GPS Carrier Phase"
14. Farrell, J., "1-sec Carrier Phase (again)"
15. Kapoor, R. S. Ramasamy, A. Gardi, R. Sabatini, “UAV Navigation Using Signals of Opportunity in Urban Environments: An Overview of Existing Methods,” 1st International Conference on Energy and Power, ICEP2016, 14-16 December 2016 (Available online here)
16. SAE International, Remote Identification and Interrogation of Unmanned Systems
17. Van Sickle, G., “GPS for military surveillance,” GPS World, Nov. 1996. 

SIDEBAR: The Advantages of Raw Measurements

A 2012 flight validation used GPS without augmenting system corrections but with raw measurements from receivers. Twenty years earlier Lincoln Labs successfully demonstrated GPS broadcasts with Mode S beacon transponders at Logan Airport as described in E. T. Bayliss et alia. (Note: the 2012 flight in the first case used UATs instead of Mode S). Transmitted positions enabled each participant to track every other participant’s data while minimizing or eliminating garble, by replacing conventional interrogations with information in assigned time slots (as ADS-B currently prescribes).

One basic modification of the Lincoln Labs methodology was advocated in the work by J. Farrell and E. McConkey and linked with another system — the Joint Tactical Information Distribution System (JTIDS) to form a more general application. Instead of coordinates, the transmitted message’s 48 information bits can contain raw uncorrected measurements. Data compression and the cycling of in-view GPS satellites can mitigate bandwidth limitations.

The introductory paragraphs of an earlier article titled “Send measurements, not coordinates,” (J. Farrell et alia) noted eight crucial advantages. In combination with each — tracking-every-other feature already noted — a later expansion of that paper (J. Farrell and M. Farrell, Additional Resources) offered an even more extensive list of advantages:

• Two decades of stunningly successful differential GPS operations demonstrate this approach
• Error source cancellation capability is intrinsic to differential GPS 
• The ability to account for specific sensitivities of each individual measurement 
• The opportunity to employ those sensitivities to assign data weighting adaptively 
• Widely known techniques for minimization of statistical error resulting from that adaptivity 
• Prompt determination of full information (cross-range as well as along range) 
• Presence in that information of accurate dynamics as well as current position 
• Ability to use the dynamics to anticipate time of closest approach 
• Ability to deduce, from the dynamics, the miss distance at that future time 
• Ability to resolve conflicts by turns or speed change instead of climb/dive 
• Applicability to both 3-D (in-air) and 2-D (runway incursion) encounters 
• Removal of potential danger in the event of datum reference non-uniformity 
• Full usage of available data when too few satellites are visible for a full fix 
• Integrity checks enabled with any number of satellites observed 
• Unrestricted algorithm release (no strings attached or proprietary claims) 
• No need for augmentation (corrections) from ground stations 
• Opportunity for participants to share observations of nonparticipants 
• Retention of applicability with or without prospective modernizations 
• Insensitivity to different models used in different constellations

These benefits are utterly absent if calculations must rely only on instantaneous position reports.

SIDEBAR: Integrity Testing: Ultra-simple and Rigorously Validated

Volumes have been written on Receiver Autonomous Integrity Monitoring (RAIM), often supported by sophisticated analytical methods and substantial mathematical development. The good news is the hard work has been done. All a program needs is a set of expressions to put into code. Even more fortuitous is further simplification of those expressions—and that also has been done. Moreover, the way that simplification has been done allows extension beyond GNSS, to include every morsel of data used for navigation.

Conventional RAIM uses five satellites for fault detection and six satellites for fault exclusion or isolation. Because every subset of four within those sets must support adequate geometric dilution of precision (GDOP), exclusion or isolation is not always available. Then, when the five-satellite detection indicates excessive error, conventional RAIM rejects the whole quintet, the good along with the bad. Forcing valid data to suffer from “guilt-by-association” is extremely wasteful and unnecessary. See here. Reversing the loss is especially urgent when data availability is marginal. A variety of advanced integrity features offers:

• Addition of cyclic bias estimation without changing navigation solutions
• Circumvention of parity vector operations added for conventional fault isolation/ exclusion 
• Replacement of that parity vector by a parity scalar with no loss of capability 
• Normalization of that parity scalar to a form with variance equal to one (dimensionless) 
• Accounting for effects of correlations incurred by differencing 
• Inclusion of closed form matrix solutions for fault detection and isolation/ exclusion with correlations 
• Extension to separate validation of each individual measurement, whether others are present or not 
• Opportunity to verify single-measurement tests when multi-satellite isolation/ exclusion is feasible 
• Support by rigorous theory (matrix decomposition etc.) with no need to employ it in operation. 
• The normalized parity scalar test for every individual measurement (everyone understands a dimensionless scalar random variable with sigma = 1) provides a vital means of operating with any and every available source of navigation information.

SIDEBAR: Double Differencing

Ignoring wide variations in sensitivity and correlation parameters is ruinous, but it is possible to recover that information.

In August 2000 I presented the raw measurements-in-squitter-messages concept as a natural extension of GPS double differencing and asked RTCA SC186WG4 members to imagine two happenings:

• Let every system and every plan in existence be only supplemental/ backup
• Let every participant compare his own data from each separate satellite to corresponding measurements from all other participants, weighting every individual difference adaptively according to its information content (we’ve been optimizing partial information weights for a half century).

A rock solid track record supports double differencing and all modes of tracking (air-to-air, air-to- surface, surface-to-air, surface-to-surface) by modern estimation. A sequence of position reports can suffice for transoceanic flight but not within crowded airspace. As noted in [4] computerized “bookkeeping” can easily maintain track files for every participant in any scenario.    

The post Enabling Collision Avoidance with Raw Measurements and Updated ADS-B Software appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

Motivation and Background

Motivation and Background
GNSS technology has become a crucial component for modern society. One of the major factors that impacts the performance of GNSS at high latitudes and equatorial areas is ionospheric scintillation. Ionospheric scintillation is usually manifested by rapid and random fluctuations in signal amplitude and carrier phase (see S. Basu et alia (2002) and Y. Jiao and Y.T. Morton (2015) in Additional Resources near the end of this article). During severe ionospheric scintillation, deep amplitude fading and high carrier dynamics may result in increased carrier tracking loop errors, carrier phase cycle slips, and potential loss of lock of signals. (see S. Skone et alia (2001) and J. Seo et alia (2011), Additional Resources). As a result, there is a need to monitor and detect ionospheric scintillation, to gain an understanding of the signal characteristics during scintillation, and to develop robust receiver algorithms that can mitigate scintillation effects.

The motivation for this project was initiated by the need to effectively and efficiently collect sufficient volume of high quality, real scintillation data for the purpose of analyzing and characterizing the ionospheric scintillation phenomena. Since 2011, the authors’ research group has been developing multi-GNSS data collection systems to capture scintillation signals by deploying these systems at various locations known to be susceptible to scintillation activities around the world (see Y. Morton et alia (2015)). Figure 1 (see figure at the top of this article) shows the locations of the established and planned data collection sites. Some of the data collection systems are equipped with devices to collect raw intermediate-frequency (IF) samples during scintillation for post-processing. This capability is especially important when commercial ionospheric scintillation monitoring (ISM) receivers are unable to maintain lock of the signals or are incorrectly estimating signal parameters during strong scintillation. A caveat of IF data collection is that it requires an enormous storage space due to the high sampling rate. Therefore, an effective scintillation detector is needed to automatically and accurately detect scintillation events and trigger the IF data collection process. The detector is also useful for researchers to sort out scintillation events collected by continuously operating ISM receivers for analysis.

Most previous scintillation monitoring and detection methods were based on commonly used indicators such as the amplitude and carrier phase scintillation indices (e.g. S₄ and σ_ϕ), and their probability density functions (PDFs) (see W. Fu et alia (1999), S. Taylor et alia (2012) and D.V. Ratnam et alia (2015), Additional Resources). These indicators are the so-called lower-order moments of the scintillation statistics derived from standard deviations of signal parameters around their nominal trends. Abnormalities in these lower-order moments due to scintillation are often indistinguishable from other effects such as multipath and interference. In addition, these previous scintillation detection methods are usually based on traditional Neyman-Pearson detection theory (S.M. Kay (1998), Additional Resources) in which certain PDFs (e.g. Gaussian) under different hypotheses have to be assumed before training and detection. For these reasons, it is difficult for these previous detectors to have reasonable false alarm rates and missed event detection rates.

While missing event detection will clearly lead to missed opportunities to study potential interesting cases, a high false alarm rate will result in the data collection system capturing events that are not of interest to the researcher and lead to wasted storage space and analysis time. 

To overcome the above problems, we have developed a new automatic scintillation detection technique using a machine learning algorithm, called support vector machine (SVM) (see articles from Y. Jiao et alia (2017, 2016 and 2017), Additional Resources). The SVM algorithm is based on the Structural Risk Minimization (SRM) principle, which seeks the boundary with the greatest separation of the two classes in the data samples (see S. Haykin, (2009), Additional Resources). The data samples are transformed into high-dimensional space, where the data will reveal features that could not be captured by lower-order moments of the signal statistics. Moreover, unlike algorithms based on the traditional Empirical Risk Minimization (ERM) (e.g. minimum square error, least squares, and least mean squares), which uses empirical PDFs of the signals to minimize the error between the desired output and the actual output (see again S. Haykin, (2009), the SVM algorithm does not have to assume the PDFs for signals under different hypotheses. Finally, the SVM algorithm can further transfer the data samples that are not originally linearly separable into even higher dimensional space where they may be linearly separable.

The basic concept of SVM is shown in Figure 2 (see figure at the top of this article). Given a training data set 

{x_p; d_p}^P_p=1 

where x_p ∈ ℝ^N is the p^th input sample, d_p = ±1 represents the desired label for two classes. If the classes are linearly separable in ℝ^N, then the discriminant function g(x) = w^Tx + b exists such that 

x_p ∈ C₁ → g(x_p) ≥ 1, then d_p = +1     (1) 

x_p ∈ C₀ → g(x_p) ≤ 1, then d_p = −1     (2)  

The margin of separation between the two classes can then be derived as 

Equation (3) (See inset photo, above right, for equations 3 & 4)

In the SVM algorithm, the goal is to maximize this margin of separation, subject to equations (1) and (2). As a result, the Lagrangian cost function in SVM can be constructed as follows: 

Equation (4) 

where α_p’s are the Lagrangian multipliers.

The further development of SVM techniques is mainly established on the solution of equation (4). For detailed mathematical descriptions and derivations, the readers are referred to Y. Jiao et alia (2017) and S. Haykin, (2009). 

In this article, we summarize the findings reported in Y. Jiao et alia (2017, 2016 and 2017), Additional Resources, which used a large volume of real scintillation data collected from stations in the northern auroral and equatorial areas to train and test the SVM-based amplitude and phase scintillation detectors. The performance of the detectors will be mainly reviewed, including validation accuracy, testing performance on novel data, and concurrent amplitude and phase scintillation detection performance using similar SVM techniques using data from equatorial regions.

Training and Validation 
All the training data was collected by commercial ISM receivers, which output 50 hertz signal intensity and 100 hertz phase measurements. Only GPS L1C/A data is used in this article with an elevation mask of 30˚ to reduce multipath effect. For amplitude scintillation detection, a total of 46 hours of data consisting of 15 segments from Ascension Island and Hong Kong are selected for training. For phase scintillation detection, a total of 28 hours of data consisting of 30 segments from Gakona, Alaska are used for training (Figure 3). The training data are partitioned into three-minute blocks. Empirical class labels are assigned to the training data based on manual inspection of the values of amplitude scintillation index S₄ and phase scintillation index σ_ϕ within each block (examples shown in Fig. 2). Only two class labels are assigned: 0 for non-scintillation data, and 1 for scintillation data. More detailed information of the training data set is listed in Y. Jiao et alia (2016 and 2017), Additional Resources.

The content of the training vector for a three-minute data block is listed in Table 1. The first entry in a column training vector is the class label assigned manually. The second and third entries are the maximum and the average S₄/σ_ϕ index values within the block. To test the impact of the index values on the performance of the SVM detectors, the second and third entries can be turned on or off in the training. The rest of the entries in the training vector are power spectrum densities (PSD) for different frequencies obtained from performing short-time Fourier Transform (STFT) on raw signal intensity and detrended phase measurements for amplitude and phase scintillation detection, respectively. 

Validation can be performed using the training data and the manually-assigned class labels. The validation methods used for amplitude and phase scintillation detection are 25% hold-out validation and 5-fold cross-validation, respectively, as described by S. Haykin, (2009). In 25% hold-out validation, 75% of the training data is selected randomly to train the detector, while the rest is reserved for validation. In 5-fold cross-validation, the training data is randomly portioned into five subsets of equal sizes. One out of the five subsets is retained as the validation data to test the model that is trained by the remaining four subsets. Then, this cross-validation process is repeated five times, so that each of the subsets is used exactly once as the validation data set. The final validation performance is an average of the five validation results. The latter validation method is more suitable for a small training data set. 

The performance of the amplitude and phase scintillation detectors is evaluated in terms of the overall accuracy, and the true positive rate (TPR) and the false positive rate (FPR) of the operating point as listed in Table 2. TPR and FPR are also commonly referred to as “hit rate” and “false alarm rate”, respectively.

They describe the probability of categorizing a target (e.g., a scintillation event) as present when it is truly present or truly absent. In general, the higher the TPR or the lower the FPR is, the better the performance of the detector is. In this study, there are four variations of the detector implementation: the S/σ_4ϕ  features (2^nd and 3^rd entries) are either included or excluded in the training vectors; and the SVM algorithm is either linear SVM or medium Gaussian kernel SVM with a kernel scale of 9.1.

Table 2 shows that both the SVM amplitude detector and phase detector have good performances. Compared to the phase scintillation detector, the general performance of the amplitude scintillation detector seems to be slightly better with a higher overall accuracy and lower FPR. The four variations of either detector show comparable performance. This indicates that non-scintillation and scintillation events are almost linearly separable in the high-dimensional space, and excluding index features from the training vectors does not influence the validation performance of the amplitude or phase scintillation detector. 

Test Performance on Novel Data
To test the generalization capability of the SVM scintillation detectors, several segments of novel data from the training data sites and from other data sites are used. Figure 4 and Figure 5 show test results of the SVM amplitude and phase scintillation detectors using novel data from the training data sites (subplots (a)) and sites not involved in training (subplots (b)). Only the linear SVM technique is used for testing as it shows similar performance to the medium Gaussian SVM in the validation, and is easier to implement.

Results in Figure 4 and Figure 5 show that all the SVM detectors are able to capture strong scintillation events. However, for phase scintillation detection, the SVM detector trained with σ_ϕ features show obvious miss-detection of medium to weak scintillation events, while the detector trained without σ_ϕ features seem to have no problem in detecting weak to strong scintillation. This phenomenon shows that the absolute values of S₄ and σ_ϕ indices alone are not reliable indicators of scintillation activity. For phase scintillation, higher dimension features such as the spectral contents may offer a more reliable means to distinguish scintillation from other activities that impact phase measurements. This is the main reason why a machining learning-based approach that exploits the high-dimensional features can outperform traditional Neyman-Pearson detectors which are solely based on assumed models of low-order statistics such as scintillation indices. 

In addition, Figure 4 and Figure 5 demonstrate that the linear SVM detectors trained without S₄/σ_ϕ features have good generalization capabilities, as they are effective on novel data taken at different locations from the training data sites.

Concurrent Phase and Amplitude Scintillation Detection at Low Latitudes
Other than the stand-alone performance of the amplitude and phase scintillation detectors, it is also interesting to investigate the relationships in SVM detector performances for both amplitude and phase scintillation on the same data set from low latitude areas, where the strongest scintillation events tend to occur. Unlike high-latitude scintillation, which is dominated by phase scintillation, scintillation observed in the low-latitude area often features concurrent amplitude fading and rapid phase fluctuations (see articles from Y. Jiao et alia (2015 and 2013), Additional Resources). Using the data from low-latitude stations, we are able to investigate this feature of low-latitude scintillation from the perspective of detection performance. To ensure that we make a reasonable comparison, the linear SVM detectors trained without S₄/σ_ϕ features are used for amplitude scintillation detection and phase scintillation detection, respectively.

Figure 6 shows results for concurrent amplitude and phase scintillation detection using SVM on novel data from Hong Kong, Jicamarca (Peru), and Singapore. The results show that the SVM phase scintillation detector trained with data from Gakona, AK can effectively operate on data from the low-latitude area. This indicates that the higher dimensional features in phase scintillation are similar in high and low latitude events.

Based on visual inspection of the values of S₄ and σ_ϕ indices in Figure 6, the two indices are highly correlated at low latitudes. However, amplitude scintillation and phase scintillation detections are not concurrent. Results show that amplitude scintillation appears to be detected more often than phase scintillation. 

To further quantify the relationship between concurrent amplitude and phase scintillation detection, Figure 7a plots the percentage of positive phase scintillation detection during positive amplitude scintillation, with respect to different amplitude scintillation levels represented by the mean S₄ values within its three-minute block. The data used for these statistics are 15 segments of novel data with total length of 53 hours from Hong Kong, Jicamarca, and Singapore. Figure 7a shows that the percentage of phase scintillation detection increases as the amplitude scintillation level becomes stronger. When the average S₄ index within a block exceeds 0.3, a concurrent phase scintillation event will definitely be detected if an amplitude scintillation event is detected. A reverse study has also been conducted and the results are shown in Figure 7b where a positive phase scintillation detection is nearly always accompanied by a positive amplitude scintillation detection.

The above relationships indicate that at low latitudes, an amplitude scintillation detector alone is sufficient to detect scintillation activities. Low level amplitude scintillation may not be accompanied by noticeable phase scintillation. However, all phase scintillations are associated with amplitude scintillations. This observation is important for low-latitude scintillation monitoring because signal intensity measurements are more reliable than phase measurements at low latitudes. For high latitudes, phase scintillation detector is needed because phase scintillation is the dominating activity as described by Y. Jiao and Y.T. Morton (2015).

Summary and Conclusions
This article introduces a SVM-based machine learning algorithm for autonomous ionospheric amplitude and phase scintillation detection on GPS signals. The input of the detectors is the PSD of the raw signal intensity and the detrended phase measurements. Instead of having to acquire knowledge of the PDFs of the signals, the machine learning algorithm learns the different high-dimensional features for non-scintillation and scintillation events from the training data, and automatically generates a discriminative hyperplane to optimally separate the two classes with the maximum separation space. 

The trained SVM amplitude and phase scintillation detectors were evaluated in validation and testing, which demonstrate good validation performance and generalization capability in testing. A summary of the findings and conclusions in this work is recapitulated below:

• The overall accuracies in the validation are around 98% and 92% for the SVM amplitude scintillation detector and phase scintillation detector, respectively.
• Linear and medium Gaussian kernel SVM perform similarly for scintillation detection.
• Excluding S₄/σ_ϕ features in the training vector does not affect the validation performance.
• Testing on novel data reveals miss-detection of weak to moderate phase scintillation events using the SVM phase detector trained with σ_ϕ features. This result shows that phase scintillation index values may not be a good indicator of the scintillation activity. Future development of the phase scintillation detector should mainly be based on features in the frequency domain, instead of the absolute values of phase fluctuations.
• The SVM detectors can be expanded to work for data from other sites not involved in training. 
• The detection of amplitude and phase scintillation may not be simultaneous with similarly implemented SVM detection techniques. At low latitudes, whenever phase scintillation is detected, it is almost certain that amplitude scintillation will be detected at the same time. On the other hand, when amplitude scintillation is detected, phase scintillation may not be simultaneously detected but the likelihood increases as scintillation intensifies.

Acknowledgement
The data collection systems are developed, assembled, and managed by CSU GPS Lab engineers Steve Taylor and Harrison Bourne. The authors wish to thank Dr. Don Hampton at the University Alaska Fairbanks, Mr. Marty Karjala at HAARP, Mr. Kevin Abnett at Poker Flat Research Range, Dr. Ding Yu Heh at the Nanyang Technological University, Dr. Marco Milla at the Jicamarca Radio Observatory, and Dr. Zhizhao Liu at the Hong Kong Polytechnic University for their support and hosting of the GNSS data collection systems. Dr. Todd Pedersen from Air Force Research Laboratory at Kirkland AFB helped to collect the GNSS scintillation data on Ascension Island. Jicamarca Radio Observatory is a facility of the Instituto Geofisico del Peru operated with support from NSF grant AGS-0905448 through Cornell University. Ms. Yu Jiao’s work is funded through a startup grant from Colorado State University and a grant from AFOSR (FA9550-14-1-0265). The data collection systems were deployed with funding support from AFRL (FA8650-08-D-1451), AFOSR (FA9550-10-1-0346), the Consortium of Ohio Universities on Navigation and Timekeeping (COUNT), and NSF (AGS-1428042).

Note: The methodologies and results presented in this article are based on materials presented in articles by Y. Jiao et alia (2017, 2016 and 2017), Additional Resources.

Additional Resources
[1] S. Basu, K.M. Groves, S. Basu and P. Sultana, “Specification and forcasting of scintillations in communication and navigation links: current status and future plans,” J. Atmo. Solar-Terr. Phy., vol. 64, no. 16, pp. 1745-1754, Nov. 2002. 
[2] S. Haykin, Neural networks and learning machines, 3rd ed., Upper Saddle River, NJ: Pearson Education Inc., 2009
[3] W. Fu, S. Han, C. Rizos, M. Knight and A. Finn, “Real-time ionospheric scintillation monitoring,” in Proc. ION GPS, Nashville, TN, 1999. 
[4] Y. Jiao and Y.T. Morton, “Comparison of the effect of high-latitude and equatorial ionospheric scintilltaion on GPS signals during the maximum of solar cycle 24,” Radio Sci., vol. 50, no. 9, pp. 886-903, Sept. 2015.
[5] Y. Jiao, J.J. Hall and Y.T. Morton, “Automatic equatorial GPS amplitude scintillation detection using machine learning,” IEEE Trans. Trans. Aerosp. Electron. Syst., 2017.
[6] Y. Jiao, J.J. Hall and Y.T. Morton, “Performance evaluations of an equatorial GPS amplitude scintillation detector using a machine learning algorithm,” in Proc. ION GNSS+ 2016, Portland, OR, 2016.
[7] Y. Jiao, J. J. Hall and Y.T. Morton, “Performance evaluation of an automatic GPS ionospheric phase scintillation detector using a machine learning algorithm,” NAVIGATION, Journal of the Institute of Navigation, 2017.
[8] Y. Jiao, Y.T. Morton, S. Taylor and W. Pelgrum, “Characterization of high latitude ionospheric scintillation of GPS signals,” Radio Sci., vol. 48, no. 6, pp. 698-708, Dec. 2013.
[9] S.M. Kay, Fundamentals of statistical signal processing: Detection theory, vol. 2, Upper Saddle River, NJ: Prentice-Hall, 1998.
[10] Y. Morton, Y. Jiao and S. Taylor, “High-latitude and equatorial ionospheric scintillation based on an event-driven multi-GNSS data collection system,” in Proc. Ionospheric Effects Sym., Alexandria, VA, 2015
[11] D.V. Ratnam, G. Sivavaraprasad and J. Lee, “Automatic ionospheric scintillation detector for global navigation satellite system receivers,” IET Radar Sonar Navig., vol. 9, no. 6, pp. 702-711, 2015.
[12] J. Seo, T. Walter and P. Enge, “Availability impact on GPS aviation due to strong ionospheric scintillation,” IEEE Trans. Aerosp. Electron. Syst., vol. 47, no. 3, pp. 1963-1973, 2011.
[13] S. Skone, K. Knudsen and M. de Jong, “Limitations of GPS receiver tracking performance under ionospheric scintillation conditions,” Phys. Chem. Earth (A), vol. 26, no. 6-8, pp. 613-621, 2001.
[14] S. Taylor, Y. Morton, Y. Jiao, J. Triplett and W. Pelgrum, “An improved ionosphere scintillation event detection and automatic trigger for GNSS data collection systems,” in Proc. ION ITM, Newport Beach, CA, 2012.

The post Automatic GPS Ionospheric Amplitude and Phase Scintillation Detectors appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.