August 2022 - Top 5 Cited Articles in Microwave engineering
Soutenance Ouzeau
1. Laboratoire de
Ecole Nationale Traitement du Signal et
de l’Aviation civile des Télécommunications
Degraded Modes Resulting from
the Multi Constellation Use of GNSS
Christophe OUZEAU
Ph.D. Defense
1 /47
2. Ecole Nationale
Laboratoire de
Traitement du Signal et des
Context
de l’Aviation civile Télécommunications
Context
• Multiplication of satellite radio navigation systems (Global Navigation
Satellite System: GNSS), the variety of radio navigation signals increases
• Global Positioning System (GPS) provides an accurate positioning service but
its standalone use cannot meet the civil aviation requirements
• The GPS is modernized progressively with new signals transmitted by new
satellites (GPS block II-R, II-F and III)
• Galileo is the European positioning system and will be operational in the next
years
• Galileo E1, E5a/E5b and GPS L1 C/A, L1C, L5 signals in Aeronautical Radio
Navigation Services (ARNS) frequency bands, interest for civil aviation
2 /47
3. Ecole Nationale
Laboratoire de
Traitement du Signal et des
Context
de l’Aviation civile Télécommunications
Context
• The EURopean Organization for Civil Aviation Equipment provides a European
forum for resolving technical problems with electronic equipment for air
transport
• The EUROCAE deals with aviation standardization and organizes Working
Groups , in particular, the WG 62 (Galileo) objectives are to*:
– Make recommendations to the Galileo project on issues of concern to civil
aviation airborne and ground equipment
– Produce a list of working assumptions for the operational use of Global
Navigation Satellite System
– Produce a Minimum Operational Performance Standard for airborne
GPS/Galileo/Satellite Based Augmentation System receiver equipment
– Produce a MOPS for both ground and airborne equipment for precision approach
– Address the need for standardisation associated with the introduction of dual
frequency Satellite Based Augmentation System services
*Terms Of Reference approved by EUROCAE council on July 8th, 2003
3 /47
4. Ecole Nationale
Laboratoire de
Traitement du Signal et des
Introduction
de l’Aviation civile Télécommunications
Introduction
• This thesis was conducted in coordination with the WG 62 and focuses on the
multi-system and multiple frequency issues of satellite navigation in aviation
applications
• We propose a combined receiver architecture and we look for algorithms
performances for civil aviation application, we have to consider standardized
assumptions and comply with International Civil Aviation Organization (ICAO)
requirements
• We focus on the interferences detection and the particular case of ionospheric
code delay estimation, when a frequency is lost because of a jammer
4 /47
5. Ecole Nationale
Laboratoire de
Traitement du Signal et des
Outline
de l’Aviation civile Télécommunications
Outline
1. GNSS applied to civil aviation operations
2. Combined receiver architecture
3. Interference detection
4. Ionospheric code delay estimation
5. Conclusion and future works
1. GNSS applied to civil aviation operations
2. Combined receivers architecture
3. Interference detection
4. Ionospheric code delay estimation
5. Conclusion and future works
5 /47
6. Laboratoire de 1. GNSS applied to civil aviation
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 1.1. SIS performance requirements
Signal In Space performance requirements
• Future combined receivers will have to comply with the following ICAO requirements:
Accuracy Accuracy Time Vertical
Typical Horizontal
Horizonta Vertical Integrity To Alert Continuity Availability
Operation Alert limit
l 95% 95% risk Alert limit
3.7 km 1×10-4/h to 0.99 to
En-route N/A 1-1× 10-7 /h 5 min 7.4 km N/A
(2.0 NM) 1×10-8/h 0.99999
En-route, 0.74 km 1×10-4/h to 0.99 to
N/A 1-1 ×10-7 /h 15 s 3.7 km N/A
Terminal (0.4 NM) 1×10-8/h 0.99999
Initial
approach
220 m 1×10-4/h to 0.99 to
Intermediate, N/A 1-1× 10-7 /h 10 s 556 m N/A -8/h
(720 ft) 1×10 0.99999
NPA,
Departure
16 m 20 m 1-2× 10-7 in
1-8×10-8 /h 0.99 to
APV I (52 ft) (66 ft) any 10 s 40 m 50 m
per 15 s 0.99999
approach
8m 1-2× 10-7 in
16 m 1-8×10-8/h 0.99 to
APV II (26 ft) any 6s 40 m 20 m
(52 ft) per 15 s 0.99999
approach
6 m to 4m 1-2× 10-7 in
Category I 16 m 15.0 m 1-8×10-8/h 0.99 to
(20ft to any 6s 40 m
Precision App (52 ft) to 10 m per 15 s 0.99999
13ft) approach
ICAO SIS performance requirements Source: [ICAO, 2006] 6 /47
7. Ecole Nationale
Laboratoire de
Traitement du Signal et des
1. GNSS applied to civil aviation
de l’Aviation civile Télécommunications
1.2. Modes of operation
Modes of operation
• Aircraft modes of operation are defined in [EUROCAE, 2007]:
– Nominal mode of operation: the receiver achieves the required level of
performance, using a pre-described, preferred combination of signals
– Alternate mode: the receiver achieves the same level of performance
than in the nominal mode, using alternative means or augmentations.
The receiver enters into this mode when one or several signals of the
nominal mode are not available
– Degraded mode: the receiver is unable to achieve the level of
performance of the nominal mode. In this case, an alert must be flagged
7 /47
8. Ecole Nationale
Laboratoire de
Traitement du Signal et des
1. GNSS applied to civil aviation
de l’Aviation civile Télécommunications
1.2. Modes of operation
Modes of operation and GNSS components (1/2)
• The WG 62 identified promising GNSS components combinations as nominal,
alternate and degraded means to provide navigation solution and integrity to
the aircraft
• We focused on the APV I phase of flight because:
– It requires vertical guidance
– It has more restrictive requirements than En-route down to NPA
Typical Nominal Alternate Degraded
Operation
•GPS Single Frequency + SBAS
En-route •Galileo Safety of Life •Galileo Single Frequency +
down to •Galileo E1/ E5b + SBAS Safety of Life No integrity information
NPA •GPS L1/L5 + SBAS •Combination of all available
pseudo ranges + RAIM
•Galileo Single Frequency +
•Galileo Safety of Life •GPS Single Frequency + SBAS Safety of Life
APV I •Galileo E1/E5b + SBAS •Galileo Single Frequency + •Combination of all available
•GPS L1/L5 + SBAS SBAS pseudo ranges + RAIM
Identified nominal, alternate and degraded modes GNSS combinations for En-Route to NPA and APV I
phases of flight Source: ConOps [EUROCAE, 2008] 8 /47
9. Ecole Nationale
Laboratoire de
Traitement du Signal et des
1. GNSS applied to civil aviation
de l’Aviation civile Télécommunications
1.2. Modes of operation
Modes of operation and GNSS components (2/2)
• The Galileo Safety of Life service (E1/E5b) satisfies needs for safety critical users and is
compliant with civil aviation applications. Integrity provided in the I/NAV message
(satellites clock and ephemeris deviation) [EUROCAE, 2007], computation of the integrity
risk at the alert limit, no protection level computation
• The Satellite Based Augmentation System is used to provide ephemeris + clock +
ionospheric corrections + DON’T USE flags to calculate protection levels (GPS SBAS)
• The Receiver Autonomous Integrity Monitoring algorithm is used to provide integrity
when GPS SBAS and Galileo SoL are not available
Typical Nominal Alternate Degraded
Operation
•GPS Single Frequency + SBAS
En-route •Galileo Safety of Life •Galileo Single Frequency +
down to •Galileo E1/ E5b + SBAS Safety of Life No integrity information
NPA •GPS L1/L5 + SBAS •Combination of all available
pseudo ranges + RAIM
•Galileo Single Frequency +
•Galileo Safety of Life •GPS Single Frequency + SBAS Safety of Life
APV I
•Galileo E1/E5b + SBAS •Galileo Single Frequency + •Combination of all available
•GPS L1/L5 + SBAS SBAS pseudo ranges + RAIM
Identified nominal, alternate and degraded modes GNSS combinations for En-Route to NPA and APV I
phases of flight Source: [EUROCAE, 2008] 9 /47
10. Ecole Nationale
Laboratoire de
Traitement du Signal et des
2. Combined receiver architecture
de l’Aviation civile Télécommunications
Onboard combined receivers architecture (1/2)
• We proposed the following receiver architecture to the WG 62, for each mode of operation:
– Navigation function selects the GNSS components combinations and provides navigation solution and
integrity
• Protection levels calculation by augmentation systems: GBAS, SBAS, ABAS (RAIM)
• Fault Detection and Exclusion : alerts
• Integrity risk for Galileo SoL
– Detection function monitors degradations at different levels within the receiver
GNSS combination The detection function is
selected not an integrity monitoring
Navigation Detection
function: function: function, it monitors
performances degradation
Selection of the Detection of GNSS
appropriate signal signals and
combination and integrity means
integrity method Performance level reached, loss or recovery
among all those loss or recovery of
available component
Operation mode navigation and detection functions 10 /47
11. Ecole Nationale
Laboratoire de
Traitement du Signal et des
2. Combined receiver architecture
de l’Aviation civile Télécommunications
Onboard combined receivers architecture (2/2)
• We proposed this receiver architecture to the WG 62 based on the
Begin operation:
following switching strategy
selection of the • Switches between modes of operation depends upon the availability
performance level of the GNSS components combinations
required
• When the receiver enters a degraded mode, it flags an alert
Alert
Performance
level
No nominal
Nominal modes available
Nominal mode of Nominal Alternate mode of modes available Degraded mode of
operation modes operation and alternate operation
available mode available
Navigation Navigation Navigation
No nominal No nominal and Detection
Detection Detection
modes alternate modes
available available
Nominal modes available
11 /47
12. Laboratoire de 2. Combined receiver architecture
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 2.1. Conclusion
Conclusion about combined receivers
• The proposed architecture, based on switching between GNSS components
combinations, is driven by detection functions:
– Detection algorithms must be implemented to flag a loss or recovery of
component, more precisely, to monitor if the system is compliant with the
performance needed to start or continue an operation: integrity, continuity,
accuracy and availability matched or not
– The receiver can decide to initiate a switch after a detection flag
• In case of degraded mode, the receiver must flag an alert [EUROCAE, 2007], and:
– We propose to try to maintain as long as possible some performances during the
current operation (reconfiguration of the navigator)
– Otherwise, other means must be used to continue the current phase of flight (INS
etc.)
• We look for algorithms performances for civil aviation application under standardized
assumptions, regards to parameters linked to civil aviation requirements (False alarm
rate, missed detection probability, protection levels…). In particular, we focused on
interferences detection and ionospheric code delay estimation
12 /47
13. Laboratoire de 3. Interference detection
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications
1. GNSS applied to civil aviation operations
2. Combined receivers architecture
3. Interference detection
4. Ionospheric code delay estimation
5. Conclusion and future works
Interference detection
• Aircraft embedded receiver interference environment:
– Pulsed interferences (DME/TACAN on L5, E5a/E5b, Radars on E5b) can affect
in particular GPS L5 and Galileo E5a, their mitigation is already studied in
details for the WG 62 [Bastide, 2004], [Raimondi, 2008]
– In band Continuous Waves (CW) and Narrow Band (NB) interferences can
affect all GNSS signals (even simultaneously), for instance the GPS L1 C/A and
Galileo E1 OS signals [Bastide, 2001], [Rollet, 2008]
13 /47
14. Laboratoire de 3. Interference detection
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 3.1. Context
Interference detection and combined receivers
• Future civil aviation combined receivers will be composed of filters
([EUROCAE, 2007]), for resistance to jammers (RF and IF filters). The resulting
interference threshold masks provide the characteristics of the interferences
mitigation receiver capability
• For civil aviation applications, interferences with power level below the
interference masks defined in [EUROCAE, 2007], are expected to generate
acceptable tracking errors
• But, CW can stay a certain time near highest amplitudes code spectrum lines
of the GNSS signals and generate larger tracking errors than expected by the
WG 62, signal-jammer relative Doppler shift rate, [Rollet, 2008] determined
this rate between 2.9 Hz/s and 3.1 Hz/s
• We focus on CW detection, with the maximum interference power specified
by the WG62: -155 dBW and a Doppler shift rate equal to 2 Hz/s
14 /47
15. Laboratoire de 3. Interference detection
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 3.2. Simulation assumptions
Impacted GNSS signals
• Priority is given to L1 C/A (BPSK) and E1 L1 C/A
OS/SoL (CBOC) signals, L5 and E5 major
threats are pulsed interferences (same band
than DME) already studied [EUROCAE, 2007] FC
f
-2Fc -Fc FR 2FC
• The PRN codes correlated at the receiver level: E1
Normalized correlator outputs
1
BPSK(1)
BOC(1,1)
CBOC(6,1,1/11) f
-2Fc -Fc FR FC 2FC
Normalized Correlation Function
0.5
• PSD of the materialized PRN code (black) :
• PSD of the materialization waveform (green)
0 • PSD of the PRN sequence (blue)
Characteristics GNSS signal
GPS L1 C/A Galileo E1 OS
-0.5
-1 -0.8 -0.6 -0.4 -0.2 0 0.2
Code Delay (Chips)
0.4 0.6 0.8 1
Code line PRN 6 38
Code delay (chips) Freq w.r.t L1 227 kHz 673.5 kHz
• Theoretical receiver correlation functions
Power w.r.t total -21.29 dB -28.81 dB
considering L1 C/A (BPSK) in blue, in the power
following, E1 OS is assumed as a BOC signal Characteristics of highest amplitude code lines
(red) instead of CBOC (black) for each signal, within the main lobes 15 /47
16. Laboratoire de 3. Interference detection
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 3.3. Impact of CW on signals processing
Impact of CW interference on signals processing
• A CW hitting a code spectrum line, affects the
receiver correlators outputs, code and carrier 39 dB Hz L1 C/A
tracking outputs and code-carrier smoothing
(100 s-Hatch filter)
• We used a MATLAB Rx simulator to process the
GPS L1 C/A, the PRN 6 worst code line is
impacted by – 155 dBW CW, Doppler shift rate
= 2 Hz per second. CW starts 200 seconds after
the tracking loop C/N0 at the correlator output (dBHz)
DLL: 1st order, PLL: 3rd order,
BW: 1 Hz, dot BW: 10 Hz,
product arctan
discriminator discriminator
Code tracking error (m) Phase tracking error (rad) 16 /47
17. Laboratoire de 3. Interference detection
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 3.4. Detection algorithms
Elaboration of detection techniques
L1 C/A correlator output
Normalized correlator outputs
• Because of the CW, sine waves appear on top
of the correlation peak when interference I prompt
occurs (on the I channel for instance) correlator
output
• Their amplitudes are dependent of:
– The jammer power,
– The PRN code spectrum line amplitude,
– The frequency offset between jammer and
nearest PRN code line
Correlator index
• Detection is achieved through monitoring of multiple correlators outputs (68
for GPS L1 C/A and 72 for Galileo E1 OS, from the spectrum characteristics,
code lines spacing)
• Two proposed detection algorithms tested over 1.5 106 outputs for each
correlator:
– Monitoring instantaneous FFT of correlators outputs
– Monitoring Auto Regressive model errors of all correlators time variations
17 /47
18. Laboratoire de 3. Interference detection
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 3.4. Detection algorithms: performance
evaluation
Detection algorithms performance evaluation process
• The performance evaluation process can be described in a few steps:
1. A detection criterion is defined from correlators outputs characteristics
2. Detection criterion parameters are set during a training stage without
interference under APV I phase of flight conditions (dynamics, multipath,
Doppler)
3. The detection threshold is set such that PFA < 1.6 10-5 /sample (for APV I
continuity, [ICAO, 2006])
4. Then the PMD value is determined, generating interferences, PMD = (number of
tests where the detection criterion is lower than the pre-defined threshold) /
(total number of tests)
5. The impact of non-detected interferences on tracking error at any time is then
discussed
18 /47
19. Laboratoire de 3. Interference detection
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 3.4. Detection: simulation assumptions
Simulation assumptions
• We considered the following simulation assumptions for a receiver onboard an aircraft:
• Normal aircraft maneuvers generated according to maximum dynamics specifications
([RTCA, 2006]):
Typical maximum values for normal
Dynamics parameters
aircraft manoeuvres ([EUROCAE, 2007])
Ground speed 800 Kt
Horizontal acceleration 0.58 g
Vertical acceleration 0.5 g
Total jerk 0.25 g/s
• Multipath generated each time a tracking procedure is initiated thanks to the
Aeronautical Channel model (DLR), considering a 10 degree elevation satellite in view
(Galileo satellites mask angle, [EUROCAE, 2008])
• Doppler shift between the jammer and the signal, with a Doppler shift rate of 2 Hz/s
• Received signals carrier to noise ratio at the correlator output level, according to the
received power level specified in [EUROCAE, 2008]:
GNSS signal GPS L1 C/A Galileo E1 OS
C/N0 39 dBHz 34 dBHz
19 /47
20. Laboratoire de 3. Interference detection
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 3.4. Detection algorithms: FFT model
Algorithm 1: FFT of the correlators outputs
Normalized correlator outputs
• Detection algorithm monitors the Fast Fourier
Transform of correlation outputs (snapshot),
Correlators outputs
[Bastide, 2001] for the L1 C/A, PRN
• Criterion is defined as: 6, impacted by a -
max_ fourier − mean(max_ fourier)
inst 155 dBW CW
std (max_ fourier)
• Parameters inside criterion determined
through a training simulation without
interferences and under APV I conditions Chip spacing
Number of correlator outputs
• Instantaneous maximum of the correlation Threshold
peak FFT determined at each instant during the for APV I
performance test simulation
Test distribution 20 /47
21. Laboratoire de 3. Interference detection
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 3.4. Detection algorithms: FFT model
Performances of the algorithm 1 (1/2)
• For GPS L1 C/A PRN 6 worst line, with:
• Maximum normal aircraft dynamics,
• Multipath (DLR model, elevation= 10°),
• Signal to jammer Doppler shift rate of 2 Hz/s,
• C/N0 = 39 dBHz at correlator output,
• PFA = 1.6 10-5 /sample (APV I)
PMD = 6.67 10-5
Missed detection probability
• Tracking error when CW not detected
• DLL: 1st order, BW: 1 Hz, dot product
discriminator,
• PLL: 3rd order, BW: 10 Hz, arctan discriminator
• PMD as a function of raw tracking error.
PMD * (number of times tracking error = N
meters mod 1 meter)/number of tracking
errors
N =Raw tracking error in meters 21 /47
22. Laboratoire de 3. Interference detection
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 3.4. Detection algorithms: FFT model
Performances of the algorithm 1 (2/2)
• Results for other L1 C/A PRN strong code
lines within main lobe, exemple of PRN 10
worst line, low tracking errors but same
PMD, considering:
• Maximum normal aircraft dynamics,
• Multipath (DLR model, elevation= 10°),
• Signal to jammer Doppler shift rate of 2
Hz/s,
• C/N0 = 39 dBHz at correlator output
• PFA = 1.6 10-5 /sample (APV I)
• For E1 OS, non detected raw tracking
errors never exceed 9 meters (1.2 m after
smoothing) obtained for PRN 38 worst line
(Power of lines lower than L1 C/A lines),
• Same assumptions except C/N0 = 34 dBHz
22 /47
23. Laboratoire de 3. Interference detection
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 3.4. Detection algorithms: AR model
Algorithm 2: Multichannel AR model (1/2)
• Interference implies abnormal
Normalized correlator outputs
correlators outputs time variations
• A 3 rd-order multichannel Auto
Regressive model used to monitor
simultaneously all correlators outputs,
at t [Marple, 1987]:
3
x [t] = ∑ai [k]xi [t − k]
ˆ i
k =1
(xi: ith correlator, a: AR coefficient)
Normalized correlator outputs
Chip spacing
Correlators outputs for the L1 C/A,
PRN 6, impacted by a -155 dBW CW
23 /47
24. Laboratoire de 3. Interference detection
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 3.4. Detection algorithms: AR model
Algorithm 2: Multichannel AR model (2/2)
• AR model error is determined, first • When impacting the L1 C/A PRN 6 worst
during a training stage: code line, considering:
3
e [t] = x [t] − x [t] = x [t] + ∑a0[k]x0[t − k]
i
0
i
0
ˆ i
0
ii
0
i • Maximum normal aircraft dynamics,
k =1
• Multipath (DLR model, elevation=
10°),
• And then during simulation tests, on
studied samples: • Signal to jammer Doppler shift rate of
3 2 Hz/s,
e [t] = x [t] − x [t] = x [t] + ∑ai [k]xi [t − k]
i i
ˆ i i
k =1 • C/N0 = 39 dBHz at correlator output
• Detection criterion is calculated
• PFA = 1.6 10-5 /sample
(norm of the E vector containing the
correlators outputs AR errors):
E[t] PMD = 10-5
log
E [t]
0
24 /47
25. Laboratoire de 3. Interference detection
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 3.5. Estimation and repair algorithm
Estimation and repair algorithm
• When the CW detection is successful,
we launch CW parameters estimation
thanks to a third order Prony model
• We repaired the correlators outputs (L1
C/A)
• We observed the code tracking error
(m), considering:
• DLL: 1st order, BW: 1 Hz, dot product
discriminator Code tracking error (m) as a function of time (sec)
Statistics Before After
• PLL: 3rd
order, BW: 10 Hz, arctan correction correction
discriminator Mean 19.9 m - 0.009 m
• Red plot: when the L1 C/A PRN 6 Standard
highest code spectrum line is Raw 10.5 m 1.9 m
deviation
impacted by a -155 dBW CW Maximum 45.2 m 8.6 m
• Blue plot: tracking loop output, the Mean 13.7 m 0.03 m
sine wave is removed from the Standard 5.3 m
Smoothed 0.04 m
correlator output deviation
Maximum 18.4 m 0.16 m
Raw and smoothed code tracking errors
statistics before and after correction 25 /47
26. Laboratoire de 3. Interference detection
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 3.6. Conclusions
Conclusions on interference detection (1/2)
• Simulation assumptions:
• Worst cases considered in terms of:
- Interference power under interference mask: maximum CW power -155 dBW,
- Code spectrum lines impacted (on PRN 6 for L1 C/A, on PRN 38 for E1 OS),
- Dynamics (maximum parameters as defined in [EUROCAE, 2007]),
- Multipath (low elevation angle)
- Minimum C/N0
• Algorithms proposed:
• Two detection algorithms based on multi correlators outputs monitoring:
- Computation of correlators FFT
- Multi channel AR model of correlators outputs
• Results obtained:
• FFT: PMD = 6.67 10-5, AR: PMD = 10-5 , but interference probability of occurrence
unknown, integrity risk = PMD * probability of occurrence?
• Maximum smoothed tracking error resulting from non-detected CW: 15 m (raw: 52
m) for GPS L1 C/A and 1.2 m (raw < 9 m) for Galileo E1 with FFT algorithm
• Capability of repair algorithm: std of the tracking error divided by 5
26 /47
27. Laboratoire de 3. Interference detection
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 3.6. Conclusions
Conclusions: contributions for CA receivers (2/2)
• Detection algorithms reduce integrity risk due to interferences
• When the CW was not detected we studied the impact on code tracking error
• When the CW was detected, a repair algorithm tested with good performances
• Possibility to switch to other GNSS components after detection, during APV I :
Details of a detection function for the particular case of interference detection for APV I 27 /47
28. Laboratoire de 3. Interference detection
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 3.6. Perspectives
Perspectives
• Need to determine minimum number of useful correlators without loss of
performance for each signal
• Make simulations to determine how far interference detection + repair + RAIM
provide integrity and accuracy compliant with APV I requirements
• But interference probability of occurrence unknown
• Tracking loop behavior during abnormal aircraft manoeuvres, repair algorithm
capability?
• Tests over actual measurements must be performed
• The major risk induced by the loss of frequency is due to the ionosphere if the
SBAS is not available, during a degraded mode
1. GNSS applied to civil aviation operations
2. Combined receivers architecture
3. Interference detection
4. Ionospheric code delay estimation
5. Conclusion and future works
28 /47
29. Laboratoire de 4. Ionospheric code delay estimation
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 4.1. Context
Ionospheric code delay estimation: context
Typical Nominal Alternate Degraded
Operation
•Galileo Safety of Life •GPS Single Frequency + SBAS •Galileo Single Frequency +
•Galileo E1/E5b + SBAS •Galileo Single Frequency + Safety of Life
APV I
•GPS L1/L5 + SBAS SBAS •Combination of all available
pseudo ranges + RAIM
• In a dual frequency nominal mode of operation, smoothed ionospheric-free range
measurements are used. The ionospheric error is estimated and corrected thanks to the use
of dual frequency measurements
• In an alternate mode of operation, the SBAS is used to provide the ionospheric corrections
• In the case of loss of frequency leading to a degraded mode, an estimation of the
ionospheric delay may be provided either by the Klobuchar model for GPS or the NeQuick
one in case of Galileo. But, the models only estimate part of the error ([RTCA, 2006], [GSA,
2008])
• This implies large overbounded ionospheric std values, that does not allow to support flight
operations that require vertical protection levels computation
– In the following, we propose algorithms to keep the accuracy of the dual frequency
ionospheric delay estimation compatible with APV I, in a degraded mode of operation
29 /47
30. Laboratoire de 4. Ionospheric code delay estimation
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 4.2. Introduction
Ionospheric code delay estimation: introduction
• In order to keep the accuracy of the dual frequency nominal mode for APV I, a
potential solution is the ionospheric delay estimation through the Code Minus
Carrier Divergence [NATS, 2003], indeed, the receiver outputs code and carrier
phase measurements
• But the carrier phase measurements can be affected by cycle slips, the integrity
of the CMC technique has to be evaluated
• First, we focus investigations on the integrity of the CMC technique by adding a
cycle slip detection algorithm and assessing its performance
– Availability of the detection method within Europe, for GPS and Galileo
constellations, considering maximum normal maneuvers [EUROCAE, 2007]
• Then, a Kalman filtering technique is proposed to estimate the CMC parameters
and to maintain the accuracy of dual frequency measurements
30 /47
31. Laboratoire de 4. Ionospheric code delay estimation
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 4.3. Cycle slip detection
Candidate methods for cycle slip detection
• We identified different cycle slip detection algorithms:
– Monitoring derivatives of carrier phase measurements
– Comparing smoothed and raw code pseudo ranges
– Making a phase prediction using Doppler measurements: robust against
high aircraft manoeuvres, needs Doppler measurements
• Algorithm tested (Doppler-predicted phase):
– Compute predicted phase:
• Fd: Doppler frequency
• : time delay between the previous and the current measurement
– Cycle slip is detected if:
31 /47
32. Laboratoire de 4. Ionospheric code delay estimation
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 4.3. Cycle slip detection
Cycle slip probability of occurrence calculation during APV I
TI PROBABILITY OF OCCURRENCE
• Pr[occurrence of cycle slip] = OVER 150 SECONDS
SIGNAL
NORMAL ABNORMAL
MANOEUVRES MANOEUVRES
Where: GPS L1 4 ms 1.0 10-3 9.2 10-2
C/A, 10 ms 7.5 10-4 6.1 10-2
Galileo
E1 20 ms 4.7 10-4 3.8 10-2
is determined from classical carrier tracking GPS L5, 4 ms 9.1 10-4 9.0 10-2
theory [Holmes, 1990], taking into account: Galileo 10 ms 6.8 10-4 6.0 10-3
E5a 20 ms 2.4 10-4 3.4 10-2
- The tracking loop std:
4 ms 9.1 10-4 9.0 10-2
- The integration time TI Galileo
E5b
10 ms 6.9 10-4 6.0 10-3
- The C/N0 = 30 dBHz 20 ms 2.4 10-4 3.4 10-2
• ∆t =150 seconds corresponds to an aircraft
- The loop bandwidth WL = 10Hz
total approach duration, it includes APV I
- The receiver dynamics ( )
- Jmax = 0.25 g/s (normal manoeuvers) or 0.74 g/s
(abnormal manoeuvers)
- is the signal wavelength 32 /47
33. Laboratoire de 4. Ionospheric code delay estimation
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 4.3. Cycle slip detection: requirements
Cycle slip detection and CA requirements
• The integrity risk due to cycle slip is expressed as the product of the algorithm
missed detection probability by the cycle slip probability of occurrence
– Integrity risk due to undetected cycle slips is taken as 10-8/approach:
• SIS integrity risk: 2. 10-7/approach or to manufacturer: 10-7 /approach
[RTCA, 2006]
• But risk not only allocated to cycle slips, and probability to have
abnormal dynamics
– Probability of occurrence of cycle slips for all signals and all integration
times is assumed as 10-3 over 150 s for normal manoeuvres and up to 10-1
for critical abnormal ones
Missed detection probability is taken as PMD_theory =10-5 for normal
manoeuvres and 10-6 for abnormal manoeuvres
• False alarm rate is taken as PFA = 1.6 10-5/sample from the APV I continuity
requirements ([ICAO, 2006])
33 /47
34. Laboratoire de 4. Ionospheric code delay estimation
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 4.3. Cycle slip detection: performance
evaluation
Cycle slip detector performance evaluation
• The methodology used to determine the smallest detectable bias with the
proposed detection algorithm can be summarized in a few steps:
1. Pseudo ranges measurements are generated without cycle slips. The detection
criterion is compared to varying thresholds
2. When PFA < 1.6 10-5/sample, the corresponding threshold is kept in memory
3. Then, pseudo ranges measurements are generated again with varying cycle slip
amplitudes. The missed detection probability is estimated for each amplitude
4. The experienced missed detection probabilities are compared to the
theoretically derived ones for normal (PMD_theory = 10-5) and abnormal
manoeuvres (PMD_theory = 10-6)
5. When PMD < PMD_theory, the corresponding cycle slip amplitude is recorded
as the minimum detectable error
34 /47
35. Laboratoire de 4. Ionospheric code delay estimation
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 4.3. Cycle slip detection: assumptions
Simulation assumptions
• Assumptions on pseudo ranges, worst conditions:
– Maximum dynamics defined in MOPS [EUROCAE, 2007] for normal and
abnormal aircraft manoeuvres
– Multipath: at low elevation angles (10 degrees)
– Noise: standard deviation of PLL and DLL outputs, assuming C/N0 = 30 dB Hz
– Ionosphere and troposphere by drawing successive independent random
Gaussian values multiplied by classical std models ([RTCA, 2006])
35 /47
36. Laboratoire de 4. Ionospheric code delay estimation
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 4.3. Cycle slip detection: performances
Cycle slip detector performances
Continuity requirements Integrity requirements
PMD_theory
Missed detection probability obtained as
False alarm rate obtained as a function of a function the minimum detectable cycle
the detection threshold used slip amplitude
• These smallest detectable cycle slips imply an error on position which depends on
the geometry. The availability of protection against cycle slip compatible with APV I
depends on geometry and must be computed at every second
36 /47
37. Laboratoire de 4. Ionospheric code delay estimation
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 4.3. Cycle slip detection: performances
Cycle slip detector performance: availability
• Determination of smallest detectable cycle slips for required PMD and PFA
through simulations
• Those cycle slips amplitudes were projected on the horizontal plane and
the vertical axis. When the position errors were lower than the alert limits,
the detection algorithm is declared available
• The availabilities were computed over Europe for GPS and Galileo
constellations in standalone modes
37 /47
38. Laboratoire de 4. Ionospheric code delay estimation
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 4.3. Cycle slip detection: performances
Availability of cycle slip detection over Europe
Availability of cycle slip detection over Europe, APV I,
100 %
degraded mode
• Availability estimated over
70° N
Europe, for both GPS and
Galileo constellations, during
APV I, maximum normal
dynamics
Latitude
• GPS (first map), elevation mask
angle: 5 degrees, period of
revolution: 24 hours
100 % 97,5 %
• Galileo (second map): elevation
mask angle: 10 degrees 70° N 31° N
elevation, period of revolution:
10 days
• Low availability mostly due to
vertical requirements:
Latitude
GPS: min=97%,
Galileo: min=98%
• HAL = 40 m, VAL = 50 m
98 %
31° N
-9° E Longitude 50° E 38 /47
39. Laboratoire de 4. Ionospheric code delay estimation
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 4.3. Cycle slip detection: conclusion
Conclusions on cycle slip detection
• The results obtained are promising (min availability = 97% for GPS, 98%
for Galileo), since:
– The availability must take into account the probability of falling into degraded
single frequency mode
– The performance test algorithm relies on worst case assumptions (simulated
pseudo ranges: low elevations, low C/N0)
• Continuity, integrity and availability of cycle slip detection plus Code Minus
Carrier divergence technique is evaluated
• In the following, the accuracy of ionospheric code delay estimation is
studied using a Kalman filter to estimate the CMC parameters
39 /47
40. Re
Laboratoire de 4. Ionospheric code delay estimation
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 4.4. CMC Kalman filtering
Code Minus Carrier Divergence and Kalman filtering
• Code ( P ) minus carrier phase( φ ) allows to estimate ionospheric code delay :
x x
Pxk − φ xk = 2 I xk − N xk λ xk + w xk + v xk
• Where:
- N is the carrier phase ambiguity from one satellite and at a given frequency (x)
- w and v are noise and multipath coming from code (w) and carrier phase (v)
• A Kalman filter is used to estimate ionospheric delay and
Receiver ambiguities:
zenith X = ( I0 N1 N2 ... N Nb _ sat )
•Where:
IPP
• is the mean vertical ionospheric code
Receiver E delay at the Ionosphere Pierce Point,
h
Ionosphere thin • is the obliquity factor
Re shell model depending on the satellite elevation angle [RTCA, 2006]
[RTCA, 2006] • The ionospheric delay at the receiver zenith can be
expressed thanks to I0 plus South-North (A) and West-
WGS 84
East (B) gradients: , in the following
we assume A=B=0
40 /47
41. Laboratoire de 4. Ionospheric code delay estimation
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 4.4. CMC Kalman filtering: measurements
Choice of actual aircraft measurements
• To comply with actual aircraft
conditions, the Kalman filter
performance is evaluated thanks
to measurements made by L1/L2
dual frequency receiver onboard
a flying Airbus aircraft around the
Blagnac airport (Toulouse,
France) Blagnac
• 8 laps recorded, minimum
number of 7 satellites in view Toulouse
• Loss of L2 frequency, leading to
GPS L1 C/A only simulated
• Comparison between single L1
C/A Kalman estimations and
classical nominal dual (L1 C/A +
L2) estimations Aircraft path around the Blagnac airport
41 /47
42. Laboratoire de 4. Ionospheric code delay estimation
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 4.4. CMC Kalman filtering: results
Kalman filter estimations
• The zenith ionospheric code delay is Number of Mean vertical ionospheric delay
samples over all tracked satellites
estimated thanks to the Kalman filter measurements
(in red) and compared to mean dual
Method Mean Standard
frequency estimation (in green) over deviation
all satellites in view Dual 3 104 11.1 m 3.3 m
• The filter is initialized in dual frequency
frequency mode and runs after a loss Single 3 104 10.9 m 1.5 m
of L2 frequency frequency
24 m
Code delay amplitude (m)
Peaks observed due
to losses of
satellites,
results obtained
for one
particular set of
6m measurements
Time elapsed since the beginning of the simulation (sec) 42 /47
43. Laboratoire de 4. Ionospheric code delay estimation
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 4.4. CMC Kalman filtering: conclusion
Conclusions and future works on Kalman estimations
• The performance of ionospheric delay estimation has been estimated in single
frequency mode for civil aviation application
• The method used is the Kalman filtering of Code Minus Carrier measurements SF : 10.9
m, DF : 11.9 m, and the std divided by 2
• The Kalman filter must be initialized in dual frequency mode, which implies to record
dual ionospheric code delay estimations to start the filter
• Trade-off between filter observation and state confidence. In our model, the
ionosphere state noise variance larger than observation noise variance (smooth
outputs), in case of scintillation, the filter innovation would increase (under-estimated
perturbation) and the filter would take time to converge (TTA during APV I?)
1. GNSS applied to civil aviation operations
2. Combined receivers architecture
3. Interference detection
4. Ionospheric code delay estimation
5. Conclusion and future works
43 /47
44. Laboratoire de 5. Conclusions
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 5.1. Combined receivers architecture
Conclusions: combined receivers architecture (1/3)
• Original contribution: future combined receivers architecture is proposed
and discussed. A switching-based strategy between nominal, alternate
and degraded modes of operation is described
• The switching strategy depends upon the targeted operation (with or not
vertical guidance). In particular, this thesis focuses on the APV I phase of
flight
• To initiate the switches between modes of operation, detection algorithms
are implemented. The performances of such detection algorithms are
assessed through simulations
• The results obtained allow to determine whether or not the algorithms
can be applied to civil aviation operations
• Another important point assessed is how to maintain as long as possible
the levels of performance required during degraded modes
44 /47
45. Laboratoire de 5. Conclusions
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 5.2. Inteferences detection
Conclusions: interference threat during APV I (2/3)
• Focus on CW interference detection: can stay a long time near high power code
spectrum lines, maximum power in compliance with the interference masks
defined in [EUROCAE, 2007]
• Focus on the GPS L1 C/A and Galileo E1 OS signals: highest power code spectrum
lines
• Multi correlators-based algorithms, continuity-compliant, provide low PMD, under
worst conditions (multipath, C/N0, dynamics), alleviate integrity monitoring
• Integrity risk not discussed because of the lack of information about interference
probability of occurrence
• When a CW is detected:
– We propose the receiver can switch to another available GNSS combination
to continue the current operation
– Another solution can consist in estimating the CW characteristics and
removing the interference effects from the correlators outputs, promising
results
• When the CW is not detected:
– The impact on tracking loops outputs is studied
45 /47
46. Laboratoire de 5. Conclusions
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications 5.3. Ionospheric delay estimation
Conclusions: ionospheric code delay estimation (3/3)
• Method proposed: Code Minus Carrier divergence technique + Kalman + cycle
slip detector under APV I degraded mode of operation
• Original contribution: cycle slip detector integrity and continuity compliant
• The availability of this technique has been studied for GPS and Galileo
constellations
• Worst cases considered: dynamics, C/N0, multipath, atmosphere
• Expected low probability to fall into single frequency mode, to be determined
• Availability expected to be compliant with APV I requirement
• Original contribution for accuracy: CMC parameters estimated thanks to a
Kalman filter initialized in dual frequency mode, accuracy maintained
• Kalman algorithm tested over a set of actual measurements
• Need to detect ionosphere scintillations
46 /47
47. Laboratoire de 5. Future works
Ecole Nationale Traitement du Signal et des
de l’Aviation civile Télécommunications
Future works
• The objective of this study (and future investigations) is to converge
towards a final architecture of receiver for each operation in all identified
configurations of operation modes
• The detection algorithms proposed in this thesis focus on interferences
(CW) and cycle slips detection. It is of interest to combine those
algorithms with RAIM-type algorithms in future investigations to know
precisely the performance of those combined algorithms for civil aviation
use
• This thesis focuses on the detection function and not on the navigation
function. However, future works may include a complete simulator of
protection levels computation, taking into account all the components
described
47 /47