Transcript Modeling Ionospheric Spatial Threat Based on Dense

ION GNSS 2008
Savannah, GA
Sept. 16-19, 2008
Modeling Ionospheric Spatial Threat
Based on Dense Observation Datasets
for MSAS
T. Sakai, K. Matsunaga, K. Hoshinoo, ENRI
T. Walter, Stanford University
ION GNSS 16-19 Sept. 2008 - ENRI
Introduction
SLIDE 1
• The ionospheric effect is a major error source for SBAS:
– The ionospheric term is the dominant factor of protection levels;
– Necessary to develop ionosphere algorithms reducing ionospheric
component of protection levels to improve availability of vertical guidance.
• Threat model should be prepared for new algorithms:
– Any algorithms need the associate spatial threat model to ensure
overbounding residual error;
– The threat model depends upon the algorithms;
– Developed a methodology to create a spatial threat model.
• Threat models created by the proposed methodology:
– Evaluation of the current MSAS threat model;
– Some new threat models evaluated; System availability also evaluated.
ION GNSS 16-19 Sept. 2008 - ENRI
MSAS Status
SLIDE 2
• All facilities installed:
– 2 GEOs: MTSAT-1R (PRN 129) and
MTSAT-2 (PRN 137) on orbit;
– 4 domestic GMSs and 2 RMSs
(Hawaii and Australia) connected
with 2 MCSs;
– IOC WAAS software with localization.
• Successfully certified for aviation use.
• IOC service since Sept. 27, 2007;
– Certified for Enroute to NPA
operations;
– Approved for navigation use in
Japanese FIR.
Launch of MTSAT-1R (Photo: RSC)
ION GNSS 16-19 Sept. 2008 - ENRI
SLIDE 3
Position Accuracy
GPS
@Takayama (940058)
05/11/14 to 16 PRN129
MSAS
Horizontal
RMS 0.50m MAX 4.87m
GPS
@Takayama (940058)
05/11/14 to 16 PRN129
MSAS
Vertical
RMS 0.73m MAX 3.70m
ION GNSS 16-19 Sept. 2008 - ENRI
Concerns for MSAS
SLIDE 4
• The current MSAS is built on the IOC WAAS:
– As the first satellite navigation system developed by Japan, the design
tends to be conservative;
– The primary purpose is providing horizontal navigation means to aviation
users; Ionopsheric corrections may not be used;
– Achieves 100% availability of Enroute to NPA flight modes.
• The major concern for vertical
guidance is ionosphere:
– The ionospheric term is dominant factor
of protection levels;
– Necessary to reduce ionospheric term to
provide vertical guidance with
reasonable availability.
ION GNSS 16-19 Sept. 2008 - ENRI
NPA Availability
SLIDE 5
MSAS Broadcast
08/1/17 00:00-24:00
PRN129 (MTSAT-1R)
Operational Signal
100% Everywhere
Contour plot for:
NPA Availability
HAL = 556m
VAL = N/A
• 100% Availability for
Enroute thru NPA.
ION GNSS 16-19 Sept. 2008 - ENRI
APV-I Availability
SLIDE 6
MSAS Broadcast
08/1/17 00:00-24:00
PRN129 (MTSAT-1R)
Operational Signal
Contour plot for:
APV-I Availability
HAL = 40m
VAL = 50m
• Vertical guidance
cannot be provided by
the current MSAS.
ION GNSS 16-19 Sept. 2008 - ENRI
Components of VPL
SLIDE 7
VPL
Ionosphere
(5.33 sUIRE)
Clock & Orbit
(5.33 sflt)
MSAS Broadcast
06/10/17 00:00-12:00
3011 Tokyo
PRN129 (MTSAT-1R)
Test Signal
• The ionospheric term (GIVE) is dominant component of Vertical Protection Level.
ION GNSS 16-19 Sept. 2008 - ENRI
Ionosphere Term: GIVE
SLIDE 8
• Ionospheric component: GIVE:
– Uncertainty of estimated vertical ionospheric delay;
– Broadcast as 4-bit GIVEI index.
• Current algorithm: ‘Planar Fit’:
– Vertical delay is estimated as parameters of planar ionosphere model;
– GIVE is computed based on the formal variance of the estimation.
• The formal variance is inflated by:
– Rirreg: Inflation factor based on chi-square statistics handling the worst
case that the distribution of true residual errors is not well-sampled; a
function of the number of IPPs; Rirreg = 2.38 for 30 IPPs;
– Undersampled threat model: Margin for threat that the significant
structure of ionosphere is not captured by IPP samples; a function of
spatial distribution (weighted centroid) of available IPPs.
ION GNSS 16-19 Sept. 2008 - ENRI
SLIDE 9
Planar Fit and GIVE
• Developed for WAAS; MSAS employs
the same algorithm;
Vertical Delay
Cutoff Radius
• Assume ionospheric vertical delay can
be modeled as a plane;
IPP
Fit Plane
IGP
• Model parameters are estimated by
the least square fit;
• GIVE (grid ionosphere vertical error):
Uncertainty of the estimation including
spatial and temporal threats.
• GIVE Equation
Formal Sigma
Spatial Threat Model
Spatial Threat
Temporal Threat
ION GNSS 16-19 Sept. 2008 - ENRI
Ionospheric Spatial Threat
IPP for fit
User IPP
IGP
SLIDE 10
• Planar fit is performed with IPPs
(ionospheric pierce points) measured
by GMS stations;
• Local irregularities might not be
sampled by any GMS stations;
• Users might use IPPs within the local
irregularities; Potential threat of large
position error;
• MSAS must protect users against
such a condition; The spatial threat
term is added to GIVE;
Rfit
Irregularity
• Spatial threat model created based
on the historical severe ionospheric
storm data.
ION GNSS 16-19 Sept. 2008 - ENRI
SLIDE 11
Example of Spatial Threat Model
Max Residual
Threat Model
• Function of fit radius (cutoff radius) and RCM metric;
• Good and bad IPP geometries are distinguished by these two metrics;
• Resulted sundersampled is roughly between 0 and 2.5.
ION GNSS 16-19 Sept. 2008 - ENRI
The Second Metric: RCM
SLIDE 12
• RCM (Relative Centroid Metric) is
used as the second metric of the
threat model; The first one is fit
radius;
IGP
dcent
Rfit
Weighted centroid of IPPs
• RCM is the distance between the
weighted centroid of IPPs and IGP
divided by fit radius;
• Using Rfit and RCM, it is possible to
distinguish good and bad geometries
of IPP distribution, and thus reduce
undersampled threat term;
• For detail, see Ref. [11].
ION GNSS 16-19 Sept. 2008 - ENRI
Methodology: Data Deprivation
SLIDE 13
IGP
R1
R2 Rfit
Threat Model
• Removes some IPPs (shown in red) for planar fit; They become virtual users;
• Residual: difference between estimated plane and removed IPPs (virtual users);
• Tabulates residuals within the threat region (5-deg square) with respect to fit
radius and RCM; The largest residual in each cell contributes to the threat model
because it means the possible maximum residual users may experience;
• MSAS employs annular (shown above) and three-quadrant deprivation (Ref. [10]).
ION GNSS 16-19 Sept. 2008 - ENRI
Problems
SLIDE 14
• Current methodology:
– Data deprivation; Annular and three-quadrant deprivation schemes;
– Problem A: Possibility that some irregularities are not sampled in the
input datasets prepared from GMS data; Only 6 domestic for MSAS;
– Problem B: Resulted threat model seems to be too much conservative.
• Proposal 1 (Problem A): Oversampling:
– Creates spatial threat model based on dense observation datasets;
– Captures any irregularities even in severe storm conditions;
– In Japan, GEONET is available source of such a dense observation.
• Proposal 2 (Problem B): Alternative deprivation schemes:
– Malicious deprivation and missing station deprivation schemes provide
realistic conditions to be considered and avoid being over conservative.
ION GNSS 16-19 Sept. 2008 - ENRI
Datasets for Oversampling
SLIDE 15
• GEONET (GPS Earth Observation
Network):
– Operated by Geographical Survey
Institute of Japan;
– Near 1200 stations all over Japan;
– 20-30 km separation on average.
• Prepared datasets:
– Small/Large datasets are extracted
from the complete datasets;
– 6-station datasets for simulating the
current model; Domestic GMSs;
– 210-station datasets for oversampling.
(Blue triangle)
(Red circle)
6-Station Datasets
210-Station Datasets
ION GNSS 16-19 Sept. 2008 - ENRI
SLIDE 16
Oversampling
• Methodology:
– Planar fit is performed based on measurements at MSAS GMSs;
– All other measurements act as virtual users; Residuals from the
estimated plane represent potential threats;
– Threat region is sampled with a great density of measurements.
• Storm Datasets:
Set #
Period
Max Kp
Remark
1
03 / 10 / 29 – 31
Storm
2
03 / 11 / 20 – 22
9
9-
3
04 / 7 / 25 – 27
9-
Storm
4
04 / 11 / 8 – 10
9-
Strom
5
06 / 12 / 5 – 7
5
Solar Flare
Storm
ION GNSS 16-19 Sept. 2008 - ENRI
Current Threat Model
Max Residual
SLIDE 17
Threat Model (Current Model)
• The threat model created by the same method as the current MSAS.
ION GNSS 16-19 Sept. 2008 - ENRI
SLIDE 18
Unsampled Threat: Safety Model
Detected
Threat
Max Residual
Threat Model (Safety Model)
• Oversampled by 210 stations; Created model: ‘Safety Model’
• Detected some irregularities not sampled by MSAS GMSs and not
reflected to the current threat model.
ION GNSS 16-19 Sept. 2008 - ENRI
SLIDE 19
Threat Detected by Oversampling
View from MSAS GMS (6-Station Set)
Oversampling (210-Station Set)
• 6-Station Set provided only one IPP within the threat region;
• The threat was detected at the upper right corner of the threat region.
ION GNSS 16-19 Sept. 2008 - ENRI
Alternative Deprivation
SLIDE 20
• Malicious deprivation (Ref. [16]):
– If storm detector trips, remove an IPP which has the largest residual from
the plane; Repeat until storm detector does not trip;
– Compute and tabulates residuals of removed IPPs;
– The number of removed IPPs is limited up to 2 for this study.
• Missing station deprivation (Ref. [11]):
– Remove IPPs associate with a GMS; Repeat for every GMSs;
– Remove IPPs associate with a satellite; Repeat for every satellites;
– Compute and tabulates residuals of removed IPPs.
• These schemes provide realistic conditions when creating a
threat model.
ION GNSS 16-19 Sept. 2008 - ENRI
SLIDE 21
Threat Model Metrics
IGP
dcent
IGP
MSA
dmin
Rfit
RCM (Used by MSAS)
•
•
•
•
IGP
Rfit
RMD
MSA
The candidate metrics as the second metric of threat models;
Relative Centroid Metric（RCM）：Distance to centroid divided by fit radius;
Relative Minimum Distance（RMD）：Distance to the nearest IPP divided by fit radius;
Minimum Separation Angle（MSA）：Maximum angle between adjacent IPPs divided by
360 degrees.
ION GNSS 16-19 Sept. 2008 - ENRI
SLIDE 22
Threat Model (RCM)
Threat Model (RCM Model)
Performance
• Malicious and missing station deprivation; Oversampled by 210 stations;
• ‘Performance’: Relationship between data coverage and the associate
overbounding sigma value.
ION GNSS 16-19 Sept. 2008 - ENRI
SLIDE 23
Threat Model (RMD)
Threat Model (RMD Model)
• Tabulated with respect to RMD metric;
• Sigma grows up quickly; RCM seems better metric.
Performance
ION GNSS 16-19 Sept. 2008 - ENRI
SLIDE 24
Threat Model (MSA)
Threat Model (MSA Model)
Performance
• Tabulated with respect to MSA metric;
• Sigma stays below 0.7m for half of trials; The best metric among three.
ION GNSS 16-19 Sept. 2008 - ENRI
SLIDE 25
System Availability
MSAS Availability for APV-I Flight Mode
Safety Model
MSA Model (Proposed)
• Evaluated system availability with the proposed threat model of MSA metric;
• Availability is improved from safety model; However not enough for service of
vertical guidance flight modes.
ION GNSS 16-19 Sept. 2008 - ENRI
Conclusion
SLIDE 26
• Needs to develop a methodology to create threat model:
– Investigating ionosphere algorithms to improve the performance of MSAS;
– Any new algorithms need the associate spatial threat model.
• Proposed methodology to create a threat model:
– The current methodology: Data deprivation;
– Oversampling and alternative deprivation are proposed;
– Evaluated candidates of threat model metric; MSA metric works well with
the proposed methodology.
• Further investigations:
– Investigate ionospheric algorithms and operational parameters which
minimizes the associate threat model;
– Consider other candidates of threat model metric;
– Temporal variation and scintillation effects.