Professional Documents
Culture Documents
A Haider azaidi2@buffalo.edu
A Haider . Scientist State University of New York
721
improvements will be estimated through software 24 MEO + 9 GEO 30 MEO
simulations. Before the results are presented, the Galileo 180
24
180
30
constellations used for the analysis are described, 150 1 150 1
16 20
followed by a brief discussion of reliability theory. Since 23 29
Galileo will be a world wide system, the simulations are
120
8
120 10
19
conducted for users all over the globe, under benign and 90 15 90 28
22 9
18
Augument of Latitude
extreme masking conditions. 60 7 60 27
14 8
GALILEO CONSTELLATIONS 30 21 30 17
6 26
The Galileo system will consist of a constellation of either 13 7
0
20
0 16
medium earth orbiting (MEO) satellites or a combination 25
-3 0 5 -3 0 6
of MEOs and geostationary (GEO) satellites. Both of 12 15
19 24
these constellations are considered baseline -6 0 -6 0 5
4 14
configurations for Galileo (Lucas and Ludwig, 1999, and -9 0 11 -9 0 23
Wolfrum et al., 1999). Within each baseline configuration 18 4
3 13
there are various altitude, orbital plane, and satellite -1 2 0
10
-1 2 0 22
3
spacing possibilities. The following four potential Galileo -1 5 0 17 -1 5 0 12
2 21
constellations were discussed in Lucas and Ludwig 9 2
-1 8 0 -1 8 0 11
(1999) and Wolfrum et al. (1999): -1 8 0 -9 0 0 90 180 -1 8 0 -9 0 0 90 180
1) 24 MEOs in three orbital planes, with an altitude Longitude of the Assending Node
of 24,000 km, inclination of 55º, augmented with Figure 1 Galileo Constellations
three GEOs
2) Same as (1), but at an altitude of 19,500 km o
90 N
3) Same as (1), but at an altitude of 24,126 km,
with nine GEOs for world wide coverage using a
Walker 24/3/2 constellation. o
60 N
4) 30 MEOs in three planes, with an altitude of o
30 N
Latitude
24,000 km
o
Taking these different constellation configurations into 0
o
account, the two constellations given in Table 1 were 30 S
chosen for the simulations. o
60 S
Table 1 Galileo Satellite Constellations
o
# MEOs Altitude i Walker GEOs 90 S o o o o o o o
180 W 120 W 60 W 0 60 E 120 E 180 W
1 24 24,126 km 55º 24/3/2 9 Longitude
722
90 N
o 1) A Type I error occurs whenever a good
observation is rejected. The probability
associated with a Type I error is denoted α.
o
60 N 2) A Type II error occurs whenever a bad
o observation is accepted. The probability
30 N
Latitude
0 1 2 3 4 5 δ o ∗ C r̂ii δ o ∗ C lii
Number of Visible GEOs ∇i = = (3)
R ii C r̂ii
Figure 3 Number of Visible Geostationary Satellites
with a 30°° Isotropic Mask Angle This is called the Marginally Detectable Blunder (MDB).
Each observation has a different MDB since each
RELIABILITY THEORY residual’s covariance matrix ( C r̂ii ) is different. Once all
Reliability refers to the ability to detect blunders in the of the MDBs have been calculated, the impact of each
measurements and to estimate the effects of undetected MDB on the parameters is assessed separately using:
blunders on the navigation solution. Reliability can be
sub-divided into internal and external reliability. Internal δˆ = − N −1 ∗ A T ∗ C l−1 ∗ ∇ o (4)
reliability quantifies the smallest blunder that can be
where A is the design matrix
detected in each observation through statistical testing of
the least squares residuals. Once the internal reliability N = A T ∗ C -l1 ∗ A is the normal matrix
has been determined, external reliability quantifies the ∇o is a column vector containing all zero’s
impact that an undetected blunder can have on the except for the MDB in the ith position.
navigation solution.
The horizontal position error (HPE) corresponding to
In order to detect a blunder using an epoch by epoch least each MDB is calculated using δ̂ from equation (4). The
squares approach, a statistical test must be performed on
largest HPE from all of the MDBs represents the external
the residuals. Hence, redundancy must exist in order to
reliability for that epoch.
detect the blunder. An unknown blunder vector, ∇, will
bias the least squares residuals r̂ according to: For a more detailed treatment of reliability theory see
Vaníček and Krakiwsky (1986), Leick (1995), and Koch
r̂ = −C r̂ ∗ C l−1 ∗∇ = −R ∗ ∇ (1) (1999).
where C r̂ is the covariance matrix of the residuals SOFTWARE SIMULATION
C l is the covariance matrix of the observations The reliability and precision improvements obtained by
augmenting DGPS with the two Galileo constellations as
R = C r̂ C l−1 and is the redundancy matrix
well as with a height and a clock constraint were
Assuming that one blunder can occur at a time, the evaluated under isotropic masking conditions and in a
blunder vector ∇ contains only one non-zero element. constricted waterway / urban canyon.
Using local residual checking, each standardized residual
The simulations were conducted over 24 hours in 60
is tested according to:
second increments, using the GPS almanac from June 6,
2000. The following reliability parameters were used: α =
r̂i
≥n α (2) 0.1%, β = 10%, and δo = 4.57, the reliability algorithm
C r̂ ii 1−
2 assumed that the residual testing was performed epoch by
epoch using no apriori knowledge of the trajectory. DGPS
The underlying assumption is that the residuals are (28 satellites available) was augmented with the two
normally distributed, and that a blunder, while biasing the Galileo constellations, a height constraint and a clock
residual, does not change its variance. Two types of errors constraint. Taking all of these combinations into account,
can be made whenever a statistical test is performed.
723
the following three different satellite constellation The shape of the constricted waterway defines the shape
combinations were simulated: of the resulting masking profile, however, the scale of the
masking profile must still be specified. Figure 6 shows the
1) DGPS
masking profile for the North / South orientation.
2) DGPS + Galileo #1 (MEO + GEO)
3) DGPS + Galileo #2 (MEO)
N
For each satellite constellation combination, the following
four types of constraints were employed:
1) No Constraints – “N”
2) Height Constraint – “H”
3) Clock Constraint – “C” C h a n n el R o tate d
4) Both Height and Clock Constraints – “B” 1 8 0 o in 3 0 o S tep s
Each of these 12 positioning methods were simulated over
24 hours at each of the computation points shown in E
Figure 4. 118 computation points were used to represent
the world.
90oN
60oN
30oN Figure 5 Constricted Channels / Waterways
L atitu d e
0o
30oS
90
60oS 80
70
M a sk A n g le (D e g)
M ax 5 6.3
90oS o 60
180 W 120oW 60oW 0o 60oE 120oE 180oW
L o n g itud e 50
40
Figure 4 Computation Points
30
A measurement variance of 1 m2 was assumed for all
satellite observations (GPS and Galileo). This can be 20
achieved in two ways: 10
1) differential GPS and differential Galileo 0
0 45 90 135 180 225 270 315 360
2) single point positioning using dual frequency A z im u th (D e g )
civilian signals with accurate broadcast orbits
and clocks, which may be possible with the Figure 6 Mask Angle Profile
modernized GPS system and Galileo
The maximum mask angle was set to 56.3º. While this is a
Variances of 4 m2 and 1 m2 were used for the height and very realistic simulation for down town urban canyons,
clock constraints respectively. The height constraint general marine navigation in most constricted waterways
variance allows for tidal variations, swells, and waves. will experience much lower masking angles. However,
The clock constraint assumes that a good quality ovenized hydrographic surveys conducted near cliff walls will
quartz oscillator is being used. sometimes encounter these extreme masking conditions.
Thus these simulations are applicable to both marine and
The isotropic mask angle simulations were performed for
land applications.
the following five mask angles 0º, 10º, 20º, 30º, and 40º.
Isotropic Mask Angles Simulations
The constricted waterway / urban canyon simulations
were performed using the channel shown in Figure 5 If all of the isotropic mask angle simulation data were to
oriented in a North / South direction. This channel was be presented, a six dimensional figure would be required
then rotated 180o in 30o increments to simulate various (computation point latitude, computation point longitude,
channel orientations. A total of six different constricted positioning method, isotropic mask angle, time, and
waterways / urban canyons were simulated and analyzed.
724
HDOP / HPE). Instead of presenting any time series, the satellite navigation system is required to meet the
95th percentile HDOP / HPE were calculated for: availability requirements. This is not the case however for
the two higher mask angles.
1) each computation point
2) all 118 computation points batched together to For the 30º mask case, DGPS + both constraints has a
generate world values 95% HDOP of 4.8 with a corresponding availability of
71%. DGPS and constraints do not provide available
In addition to calculating the 95th percentile values, the
navigation, Galileo is required. Once DGPS is augmented
percentage of time that the HDOP ≤ 2 and the with either Galileo constellation the 95% HDOP
HPE ≤ 10 m were also calculated for the two cases listed decreases to ≤ 1.8 and the availability increases to ≥ 97%.
above. These limits were chosen since for many marine Adding both constraints increases the availability to
navigation applications the position is only considered
≥ 99%. At the highest isotropic mask case of 40º, the 95%
available if the HDOP ≤ 2, and the required positional
HDOP is > 10 with an availability is < 21% for DGPS +
accuracy is usually 10 m. The percentage of available /
both constraints. Augmenting with Galileo and both
reliable positions quoted in the following sections are in
constraints reduces the 95% HDOP to < 3.5 and increases
the context of these values.
the availability to 80%. Thus Galileo greatly improves the
Availability Results availability under the worst case masking conditions.
A summary of the isotropic mask angle availability results
for the world are given in Figure 7 and Figure 8. The
100
results are presented in graphical form with the 12
positioning methods on the x-axis grouped according to 90
the satellite constellations employed. Within each group
are the four constraints. Figure 7 plots the overall world 80
P e rc en ta g e o f H D O P s < 2
95% HDOP, while Figure 8 plots the overall world
70
percentage of HDOPs ≤ 2.
60
10 50
9 40
0 0
8 10 30 10
20 20
7 30 20 30
40 40
6 10
HD OP
5 0
N H C B N H C B N H C B
4 D G PS+ M EO+GEO M EO
3 Figure 8 Percentage of HDOPs ≤ 2, for the World
2 Isotropic Mask
725
from a low of 1.6 to a high of > 10. Unavailability bands The following statistics were generated for the 95%
are found in the 30º-75º latitudes. Although on the world HDOP values for the 118 computation points (30º mask
scale DGPS + height for the 30º mask case does not case) for each positioning method:
provide available navigation, some locations still have
1) minimum value (0%)
acceptable availabilities.
2) median value (50%)
3) 95% value (95%)
90oN 4) maximum value (100%)
Figure 11 shows the results of this analysis, the maximum
o
60 N and minimum values are the red lines with red triangle
markers. The median value is the dark green line with
30oN
Latitude
dark green “o” markers. The 95% value is the light green
0o shaded area. When DGPS is only augmented with
30oS
constraints, there are large variations between the
minimum and maximum values. Even when Galileo #1 is
60oS added, there are still variations (1.1 to 3.7). Galileo #2
tends to moderate the results reducing the maximum
values while slightly increasing the median values over
90oS o
180 W 120oW 60oW 0o 60o E 120o E 180oW Galileo #1.
Longitude
10
0 2 4 6 8 10 9
95% HDOP
8
Figure 9 HDOP 95%, DGPS + Height
30°° Mask Angle 7
Unfortunately the converse is also true, methods with 6
good world availabilities will have availability holes. For
HDOP
o
0 Reliability Results
o
30 S The reliability results will be presented in a similar
o
60 S manner to the availability results. A summary of the
isotropic mask angle reliability results for the world are
o given in Figure 12 and Figure 13, for the 95% HPE and
90 S o
the percentage of HPEs ≤ 10 respectively. DGPS with
o o o o o o
180 W 120 W 60 W 0 60 E 120 E 180 W
Longitude
constraints showed available positioning for the 0º, 10º,
and 20º masking cases. However, reliable navigation is
0 2 4 6 8 10 only possible for the 0º and 10º masking cases. For the
95% HD OP 20º mask case, unaugmented DGPS has a 95% HPE of
> 100 m with a corresponding reliability of 54%
Figure 10 HDOP 95%, DGPS + Galileo #1, (percentage of HPE ≤ 10 m). Augmenting DGPS with
30°° Mask Angle both constraints improves the 95% HPE to 21.1 m and
726
increases the reliability to 85%. This still does not meet Even the most augmented case for the 40º mask case has a
the desired reliability of ≥ 95%. However, augmenting maximum reliability of 70%. While this is a dramatic
DGPS with either Galileo constellation decreases the 95% improvement over the DGPS and constraints value of 5%,
HPE to 4.1 m and increases the reliability to 100%. it falls short of the 95% reliability goal.
At the higher masking angles of 30º and 40º, DGPS with Overall the two Galileo constellations perform almost
both constraints have 95% HPE > 100 m, with reliabilities identically for all of the mask cases. Only at the 40º mask
of 38% and 5% respectively. This is well below the case is there a noticeable difference in the 95% HPE
desired reliability level of ≥ 95%. For the 30º mask case (Figure 12) performance, with Galileo #2 outperforming
augmenting DGPS with either Galileo constellation and Galileo #1. However the reliability results (Figure 13) are
one constraint produces 95% HPE < 8 m and reliabilities again almost identical. This indicates that Galileo #1 has
> 97%. In this case reliable navigation is achieved. larger outliers than Galileo #2.
The regional differences first become apparent for the 20º
100 mask case using DGPS + height. Figure 14 is a contour
0 graph of the 95% HPE using the individual computation
90 10 point values. The overall 95% HPE is 30.9 m with a 79%
20
80 reliability. However, there are large variations in the 95%
30
HPE with locations ranging from a low of 5.9 m to a high
70 40
of > 100 m. Regions of poor reliability are found at ±60º
60 latitudes. Although on the world scale DGPS + height for
H P E (m )
10
o
0
0 o
N H C B N H C B N H C B 30 S
D G PS+ MEO+GEO MEO
o
60 S
Figure 12 95% HPE, for the World, Isotropic Mask
o
90 S o o o o o o o
100 180 W 120 W 60 W 0 60 E 120 E 180 W
Longitude
90
P e rc en ta g e o fH P E s < 1 0 m
80
0 20 40 60 80 100
70 95% HPE (m)
727
Constricted Waterway / Urban Canyon Simulation
100
90
The results for the constricted waterway / urban canyon
simulations are presented using the same format as the
80 isotropic mask angle results. For each orientation of the
70 constricted waterway / urban canyon (six orientations)
and for all of the orientations batched together the 95th
H P E (m )
60
percentile HDOP / HPE were calculated:
50
1) for each computation point
40 2) all 118 computation points batched together to
30 generate world values
20 In addition the percentage of time that the HDOP ≤ 2 and
10
the HPE ≤ 10 m were also calculated for the cases listed
M ax
M ed ia n above.
0 M in
N H C B N H C B N H C B Availability Results
D G PS + M EO+GEO MEO
A summary of the constricted waterway / urban canyon
Figure 15 Statistics for Computation Points’ availability results for the world are given in Figure 17
95% HPE, 20°° Mask Angle and Figure 18. For each positioning method the 95th
percentile / percentage is plotted for each orientation of
augmented with Galileo for this masking case there are no the constricted waterway, resulting in six data points. The
regions of poor reliability. overall result for all of the orientations is plotted as the
The regional reliability performance differences between solid magenta line through all of the positioning methods.
the two Galileo constellations become apparent at the 30º When six data points are not identifiable for a specific
mask case as shown in Figure 16. Galileo #1 has lower case, it is because several points overlap. Figure 17 plots
minimum and higher maximum 95% HPE values the overall world 95% HDOP, while Figure 18 plots the
compared with Galileo #2 (MEO only). This matches overall world percentage of HDOPs ≤ 2.
with the overall world reliability trends seen in Figure 12.
Thus Galileo #2 tends to moderate the results. 10
9
100 8
90 7
80 6
HDOP
70 5
60
HPE (m)
4
50 3
40 2
30 1
20 0
Max N H C B N H C B N H C B
10 Median D G PS+ MEO+GEO MEO
Min
0 Figure 17 95% HDOP, for the World
N H C B N H C B N H C B
DGPS+ MEO+GEO MEO Constricted Waterway / Urban Canyon
The overall results are similar to the 40º isotropic mask
Figure 16 Statistics for Computation Points’
angle case. DGPS + both constraints has a 95% HDOP
95% HPE, 30°° Mask Angle > 10 with a corresponding availability < 12%. Once
Galileo is added, the results improve to a 95% HDOP of
≤ 5.9 with an availability ≥ 53%. However, if the channel
728
is oriented in the north / south direction the 95% HDOP
10
increases to 9.6 and the availability decreases to 43%. If
one constraint is also added to the DGPS + Galileo system 9
all of the channel orientations have similar availability
results. 8
7
100
6
90
HDOP
Max
5
80
P e rc en ta g e o f H D O P < 2
4
70 Median
3 Min
60
2
50
1
40
0
30 N H C B N H C B N H C B
DGPS+ MEO+GEO MEO
20
Figure 19 Statistics for Computation Points’
10 95% HDOP, Constricted Waterway / Urban Canyon
0
N H C B N H C B N H C B
D G PS+ MEO+GEO MEO 100
Figure 18 Percentage of HDOPs ≤ 2, for the World 90
Constricted Waterway / Urban Canyon
80
In order to examine how the availability results varied
with user location, the 95% HDOP statistics for the 118 70
computation points with all of the orientations batched
60
H P E (m )
Reliability Results 0
N H C B N H C B N H C B
A summary of the constricted waterway / urban canyon D G PS+ M EO+GEO MEO
reliability results for the world are given in Figure 20 and
Figure 21. Figure 20 plots the overall world 95% HPE, Figure 20 95% HPE, for the World
while Figure 21 plots the overall world percentage of Constricted Waterway / Urban Canyon
HPEs < 10 m. In order to examine how the reliability results varied with
The reliability results are very similar to the 40º isotropic user location, the 95% HPE statistics for the
mask case. DGPS with both constraints has a reliability of 118computation points with all of the orientations batched
only 4%. By augmenting with Galileo, the reliability can together were analyzed as shown in Figure 22. When
be increased to 68% and 60% for Galileo #1 and Galileo DGPS is augmented with Galileo #1 and constraints the
#2 respectively. While Galileo #1 + both constraints has 95% HPEs ranged from a low of 10.9 m to a high of
the higher overall reliability there is a 10% reliability > 100 m with 95% of the locations being ≤ 97.4 m. When
variations between channels, while with Galileo #2 + both DGPS is augmented with Galileo #2 and constraints the
constraints there is only a 3% reliability difference 95% HPEs ranged from a low of 25.1 m, to a high of
between channels.
729
68.6 m with 95% of the locations being ≤ 56.5 m. Again is poor for the higher isotropic mask angles. While DGPS
Galileo #2 (MEO only) produces much more balanced and constraints do not provide reliable navigation for
results. masking angles ≥ 20°, once Galileo is added most of the
availability and reliability requirements can be met for
100 isotropic masking angles ≤ 30°. When extreme mask
90 angles are encountered (≥ 40° isotropic and severe
constricted waterways and urban canyons), DGPS must
P e rc en ta g e o f H P E < 1 0 m
10 REFERENCES
0 Leick, A., 1995, GPS Satellite Surveying, 2nd Edition,
N H C B N H C B N H C B
D G PS+ MEO+GEO MEO John Wiley & Sons, Inc. 1995.
Figure 21 Percentage of HPEs ≤ 10 m, for the World Lucas, R. and D. Ludwig, 1999, Galileo: System
Constricted Waterway / Urban Canyon Requirements and Architectures, Proceedings of
the ION GPS 99 Conference, Nashville,
Tennessee, September 14-17, 1999.
100 Office of the US President’s Press Secretary, 2000,
90 Statement by the President Regarding the United
States' Decision to Stop Degrading Global
80 Positioning System Accuracy, May 1, 2000.
70 M ax Koch, K.R,. 1999, Parameter Estimation and Hypothesis
Testing in Linear Models (2nd Edition),
60 Springer-Verlag, New York, 1999.
H P E (m )
730
Spilker, J. J. Jr., 1994, Satellite Constellation and
Geometric Dilution of Precision, Chapter 5 in
Global Positioning System: Theory and
Applications Volume I, ed. B. W. Parkinson and
J. J. Spilker Jr., Volume 163 Progress in
Astronautics and Aeronautics, American Institute
of Aeronautics and Astronautics, Washington,
DC, 1996.
Vaníček, P., and E.J. Krakiwsky, 1986, Geodesy: The
Concepts (2nd Edition), North-Holland,
Amsterdam, 1986.
Walker, J. G., 1978, Satellite Patterns for Continuous
Multiple Whole-Earth Coverage, IEE
Conference Publication No.: 160, pp 119-122,
1978.
Wolfrum, J., M. Healy, J.P. Provenzano, and T.
Sassorossi, 1999, Galileo - Europe’s
Contribution to the Next Generation of GNSS,
Proceedings of the ION GPS 99 Conference,
Nashville, Tennessee, September 14-17, 1999.
731