Professional Documents
Culture Documents
Shade Ion2004 Abwerzger
Shade Ion2004 Abwerzger
n
w
w
w
L
O M M
L
L
0 0
0
0 0
0 0
2
1
1
W
with
2
i i
w = .
The general formulation for least-squares adjustment
reads
( ) l W A A W A x
T T
1
'
=
where the vector x holds the estimated parameters (e.g.,
position, clock error,), A is the design matrix, and the
vector l holds the observations (e.g., GPS pseudo range
measurements). The matrix N
( ) A W A N
T
=
is called the normal equations matrix. The inverse normal
equations matrix finally is the cofactors matrix
corresponding to the estimated parameter vector x.
Assuming the unit weight factor
2
0
m being 1, the
variance/covariance matrix of the parameters reads
( )
(
(
(
(
(
= = =
2
2
2
2
1 2
0
1
T Tu Te Tn
uT u ue un
eT eu east en
nT nu ne north
d d d d
d d d d
d d d d
d d d d
m x N N
Knowing the variance/covariance matrix, the horizontal
position variance can be obtained from
2
2
2 2 2 2
2 2
ne
east north east north
hor
d
d d d d
d +
|
|
\
|
+
+
=
and the vertical position variance
2
u
d can directly be
taken from the variance/covariance matrix.
Finally, the horizontal and vertical protection levels (HPL
and VPL) are obtained from
hor Ped H Ped
d HPL =
,
and
u Ped V Ped
d VPL =
,
.
Determination of -values
As stated introductory in this chapter, the -factors need
to be re-assessed for the application requirements.
Transferring the aviation approach to the SHADE
pedestrian application, the precision approach mode is
most applicable due to the vertical accuracy and integrity
requirements in SHADE. However, for the horizontal
protection level (HPL), the argument for using a one-
dimensional distribution does not hold anymore unlike
aircrafts during precision approach, pedestrians have no
favored direction in their movements. A one-dimensional
(Gaussian) Normal distribution can be used for the
pedestrian vertical protection level (VPL), whereas the
pedestrian HPL follows a two-dimensional Rayleigh
distribution, comparable to the non-precision approach
case described in RTCA (1999) [1].
According to the user requirements identified for
SHADE, the applicable integrity risk for the pedestrian
user is 1*10
-3
per 60 seconds. According to RTCA (1999)
[1], half of the total integrity risk is allocated to protection
level bounding within the signal-in-space (SIS) and user
equipment domain. This requirement is allocated equally
to the horizontal and vertical protection level risk. This
results in equal integrity risk requirements of 2.5*10
-4
per
60 seconds for HPL and for VPL. The other half of the
total integrity risk is allocated to the "system-internal"
component. This part comprises, e.g., undetectable
ionospheric blunders, XPL formula inaccuracies, UDRE
tail effects and others (see also Flament (2004) [2]).
For deriving -values, it first must be ensured that the
probability of missed detection of a faulty position
solution associated to the XPL algorithm is expressed per
sample (i.e., per each XPL computation). To establish the
link between this probability and the given integrity
requirement, it is necessary to make an assumption on the
number of independent samples per time unit.
For the three SHADE pilots, different intervals of
statistically independent samples are applicable. In pilot 2
and 3, the sampling interval of 6 seconds for independent
samples is dominated by the update interval of EGNOS
fast corrections (carrier phase smoothing is not applied).
However, in case of total GPS outages, the position
solution implemented in Pilot 2 solely relies on dead
reckoning and barometric height determination. In this
case, the variance of subsequent samples will become a
function of time. First investigations of suitable XPL
computation algorithms have been carried out for this
special case. As a first approach the integrity risk was
defined as risk=2.5*10
-5
/ (1+t), whereas t[sec] represents
the time span since the last GPS/ EGNOS position fix.
The LORAN-C measurements of Pilot 3 have a sampling
interval of 5 seconds, thus being lower than the EGNOS
update. The A-EGNOS system of Pilot 1 has a sampling
interval of 6 seconds, which is equal to the fast
corrections update interval, and, hence, does not change
the -factors.
Table 1 summarizes the computed -factors for the
SHADE Pilots.
Table 1: Sample rates and integrity risk for the SHADE
pilots
Pilot Sample
interval
Integrity risk HPL VPL
1 6 sec
2.5*10
-5
4.60 4.21
2 6 sec /
f(t)
2.5*10
-5
/
f(t)
4.60 /
f(t)
4.21 /
f(t)
3 6 sec
2.5*10
-5
4.60 4.21
FIELD MEASUREMENTS RESULTS
Field demonstrations of the SHADE system under
contract from ESA encompass eight campaigns with
different key application objectives. They were performed
over a scheduled duration of six months in spring and
summer 2004. The environmental scenarios included the
Expo98 Park at Lisbon (Portugal), a hotel and an office
building at Bolzano (Italy) and a road tunnel in the center
of Rome (Italy). The involvement of user groups (public
safety authorities, fire brigades, etc.) into the
demonstrations aimed at proving application domain
acceptance and provided detailed feedback on the current
system and desired enhancements. A final demonstration
will be performed at ESTEC premises on Sept. 30.
Some test results regarding the navigation performance of
the different Pilots are presented below.
Pilot 1 Measurement Results
From the design point of view, the main advantage of the
A-GPS concept is the high sensitivity of the used GPS
receiver. This should enable to determine a GPS-based
position solution also in indoor environments and other
situations, characterized by GPS signal obstructions.
Figure 4 shows a scatter plot of static A-GPS
measurements (MU-mode). These measurements have
been taken outdoors, under good GPS reception
conditions. The position sampling rate was one second,
and the measurements have been taken over a time span
of circa 30 minutes. The blue dot indicates the precisely
known reference position, and the red dot represents a
mean value of all recorded position fixes. Although the
results show a slight southerly offset from the reference
position, such performance is typical for GPS single point
positioning techniques. The results can therefore be seen
as reference case for the subsequent evaluations.
Figure 4: Static A-EGNOS solution, outdoor
In contrast to the outdoor case, Figure 5 shows a scatter
plot of static A-EGNOS measurements in indoor
environment (business building with normal windows
front). As before, the position sampling rate was one
second, and the measurements have been taken over a
time span of circa 30 minutes. The blue dot indicates the
precisely known reference position, and the red dot
represents a mean value of all recorded position fixes.
The standard deviation of the indoor position fixes
amounts to 39.22 m (95% confidence level) which is three
times higher as the GPS SPS horizontal accuracy (cf. U.S.
Assistant Secretary of Defense (2001) [3]).
Figure 5: Static A-EGNOS solution, indoor
Figure 6 shows a kinematic A-EGNOS position track,
derived from measurements taken near Bolzano (Italy).
Measurement update rate was 5 seconds. The red dots
indicate the computed position solutions and the cyan line
represents the reference track. The yellow shaded area
indicates indoor measurements within an ancient castle,
and the green line marks a forested area in difficult
terrain.
During the whole kinematic test, the position solutions of
Pilot 1 are close to the actual path, except during the deep
indoor measurements.
Figure 6: Kinematic A-EGNOS solution, varying
environment
Generally it can be stated that the high sensitivity
provided by the A-GPS receiver significantly increases
the availability of a GPS position solution. Even indoors,
a position solution is available. However, indoor
environments are characterized by indirect signal
reception and according heavy multipath contamination.
This is directly reflected by the relatively poor position
accuracy within such areas.
The A-EGNOS technology is therefore suitable for
applications where high availability of the position
solution, but low accuracy is required. For the mentioned
security critical operations, the A-EGNOS technology is
only expected to meet the requirements with respect to
availability and reliability. But the accuracy requirements
of some 3 10 m (95% confidence level) for security
critical and search and rescue (SAR) applications, which
have been identified in the course of the SHADE project,
will not be met in many situations. Only the accuracy
requirement of 150 m (95%) for E112 (European
emergency call services) applications is well covered by
the accuracy performance of the A-GPS technology.
Pilot 2 Measurement Results
The SHADE Pilot 2 bases on a GPS/EGNOS receiver
augmented by autonomous sensors (magnetometer,
barometer and accelerometer) for position determination.
This technique theoretically also allows position
determination within GPS hostile environments, given an
absolute position could initially be determined by the
GPS/EGNOS receiver.
As Pilot 2 uses a COTS consumer grade GPS/EGNOS
receiver, the results presented here will focus on the
performance of the autonomous sensors, i.e., the ability of
the system to continue position computation during GPS
outages.
Figure 7 shows a comparison of Pilot 2 position solutions,
derived from different positioning modes. The thin black
straight line represents the reference trajectory, obtained
from precise geodetic measurements. The red dots
illustrate position fixes in GPS stand-alone mode. At the
end of the track (lower left corner of the figure), some
outliers caused by multipath effects can clearly be seen.
The green dots represent the integrated position solution:
GPS/EGNOS and the autonomous sensors measurements
have been merged by the Kalman Filter for position
determination. In this case the outages experienced in the
GPS/EGNOS stand-alone case could easily be
compensated by the autonomous sensors.
The blue dots finally represent the calibrated sensor
solution. GPS was only used during start-up epoch for
absolute position determination. Afterwards, the solution
solely relies on the measurements of the autonomous
sensors, i.e., it was derived in dead reckoning mode. It
can clearly be seen that those sensors are still able to
measure the shape of the track. However, the position
fixes are affected by a slight linear position drift. In this
case, the absolute position difference to the reference
trajectory at the end point amounts to circa 21 meters.
Figure 7: Kinematic Pilot 2 position solutions I
(comparison)
Figure 8 shows results of another kinematic Pilot 2 test.
The main task of this test was to investigate the
performance of the system while walking along streets
and around a building. This test was done on a parking lot
with an unobstructed view to the sky. The GPS/EGNOS
solution showed good performance with high accuracy
(~1.5m; 95%) and is therefore used as reference.
Compared to the trial shown in Figure 7, the sensor
solution is not affected by the linear position drift.
Figure 8: Kinematic Pilot 2 position solutions II
(comparison)
These kinematic trials demonstrate that the technology
used for Pilot 2 has great potential for bridging total GPS
outages and continuing positioning with GPS-like
accuracy. As the position plots show, the autonomous
sensor solution is exactly able to represent the shape of
the track. One unsolved issue connected to this solution is
the commonly occurring linear position drift being caused
by the magnetometer. The most likely reasons for these
drifts are:
Miscalibration of magnetometer
Misalignment of magnetometer with respect to
walking direction of the pedestrian
Uncalibrated influences of magnetic inclination and
declination.
Further investigations are currently carried out. Electric
installations within buildings cause distortions to the
direction measurements and thus significantly decrease
the Pilot 2 position accuracy in such environments. A
solution that uses gyro measurements for stabilizing the
magnetometer and thus can compensate temporal
distortions of the magnetometer is currently under
investigation.
The Pilot 2 technology offers great potential for meeting
all requirements, which have been identified for the
SHADE users and associated applications. First tests, also
carried out in indoor environment, show excellent results.
However this technology still needs further investigations
and developments before it can be seen as ready for the
market.
Pilot 3 Measurement Results
The SHADE Pilot 3 bases on an integrated GPS/EGNOS
and Loran-C navigation system. Due to the signal
characteristics of Loran-C, this technique theoretically
allows position determination within GPS hostile
environments. As Loran-C provides very poor absolute
accuracy but high relative accuracy and also good
repeatability for the position solution, this technique
needs a calibration strategy for Loran-C. Therefore,
during good GPS availability and quality, Loran-C time
of arrival (TOA) measurements are continuously being
calibrated. Once GPS availability is limited or if there are
even complete GPS outages, the calibrated Loran-C
measurements can be used to continue positioning with
GPS-like accuracy.
Results of previous Loran-C measurements in static and
kinematic environments have already been presented and
they have proven the theoretical potential of the concept
(e.g., Abwerzger and Lechner (2003) [4]).
In the course of the SHADE project, Pilot 3 has also been
tested in dense urban and indoor environments. Indoor
tests have been carried out inside a hotel and a car dealers
complex near Paris (France). The results of these trials
were not as expected: Upon entering the buildings, the
signal strength of all received Loran-C stations decreased
so dramatically that the Loran C receiver could no longer
reliably track the stations. Hence, a position solution
inside these buildings was not possible.
This fact indicates that the quality of Loran-C signals is
significantly lower in the vicinity of electric installations,
as well as in the vicinity of objects made of metal. Re-
radiation effects and disturbing frequencies can be hold
responsible for disturbing the Loran-C reception. At least
the used receiver could not cope with these disturbances
and ceased to output reliable Loran-C measurements.
After leaving the buildings, and after about two minutes
re-acquisition time, the Loran-C signals could again be
received with sufficient quality for position
determination.
Further tests in dense urban environment were afterwards
carried out in the city of Reims near Paris. Figure 9 shows
the measured trajectory superimposed on a city map. The
true trajectory is indicated by black dots, while red dots
indicate an integrated GPS/Loran-C position solution,
green dots indicate GPS stand-alone position fixes, and
blue dots indicate Loran-C stand alone position fixes.
Figure 10 and Figure 11 show time series of Northing and
Easting coordinates from the different positioning modes
of the same test. Here it can easily be seen how the
navigation systems complement each other and contribute
to an enhanced overall availability of the position solution
Numerical analysis showed that during 220 epochs (35 %)
less than three GPS satellites were visible, whereas only
during 83 epochs (13 %) less than 3 three Loran-C
stations were received. Expressed the other way round,
only 65% of the time a GPS only solution would have
been possible, while 87% of the time a Loran-C only
position solution would have been available. The
combined position availability however was 97 %.
Figure 9: Kinematic Pilot 3 position solutions superposed
to city map (black dots indicate true trajectory)
The kinematic Pilot 3 tests have shown that the reception
quality of Loran-C signals is very much influenced by the
immediate environment. That is, electrical installations,
power lines, or even objects made of metal can negatively
influence the Loran-C signal quality. The receiver (if the
receiver might be held responsible at all; tests are
ongoing) used for the tests could not cope with the
problems, ultimately causing unavailability of the Loran-
C measurements in these environments.
Figure 10: Northing time series in different positioning
modes
Trials within dense urban areas have shown that Loran-C
is in fact able to significantly increase the availability of
the combined position solution and to bridge outages of
GPS. The accuracy of calibrated Loran-C position
solutions is below the expectations. Furthermore, Loran-C
position solutions show unresolved systematic error
effects, which need to be investigated in more detail. As
kinematic Loran-C tests carried out within the scope of
earlier measurement campaigns do not show the same
systematic effects, it must be assumed that the Loran-C
receiver used in SHADE is responsible for those effects.
Some hints also indicate that the used notebook may have
interfered with the Loran-C receiver. These issues will be
investigated further in the near future.
Figure 11: Easting time series in different positioning
modes
The theoretical concept of Pilot 3 looks promising, but
currently there are a number of technical issues, which
remain to be solved. Consequently, the availability and
also the accuracy of the Pilot 3 position solution during
poor GPS availability and total GPS outages is still too
low for the applications envisaged in SHADE.
CONCLUSIONS
The introduction of a positioning system into safety-
critical applications remains an ambitious objective,
although the potential benefit is clear at hand. The system
architecture of SHADE was designed to improve
availability of continuous positioning solutions with
integrity information even in dense urban and indoor
environments. Feedback and acceptance of the user
groups to the system architecture and to its performance
was obtained during several field demonstrations of the
SHADE Pilots.
From a technical point of view, the following conclusions
can be drawn for the different SHADE pilots:
Pilot 1 technology is suitable for applications where
high availability of the position solution, but low
accuracy is required. Accuracy requirements for most
SHADE applications will not be met.
Pilot 2 technology showed most promising results and
is likely able to meet most requirements of the
SHADE applications. However, there are still some
technical problems to be solved.
The concept of Pilot 3 technology was promising, but
due to technical limitations, most requirements of
SHADE applications have not been met. Especially in
indoor environments the concept fails. However,
Loran-C receiver manufacturers are continuously
improving the performance of their receivers. It is
expected that the indoor performance of Loran-C will
improve in the near future.
ACKNOWLEDGMENTS
The project "Definition and Demonstration of Special
Handheld based Applications in Difficult Environment" is
performed by TeleConsult Austria and its partners under
contract to the European Space Agency (contract number
17545/03/NL/GS).
REFERENCES
[1] RTCA (1999): Minimum Operational Performance
Standards for Global Positioning System/Wide Area
Augmentation System Airborne Equipment, RTCA-DO
229 B, October 6, 1999.
[2] Flament D. (2004): EGNOS System and Performance,
Presentation to CNIG/PSD: 25/03/04, available at
http://www.esgt.cnam.fr/sites/CNIG/cnig.psd/CIAG/
CNIG.PSD/reunions/25mars2004/EGNOS_DFlament.PD
F (July 2004)
[3] U.S. Assistant Secretary of Defense (2001): Global
Positioning System Standard Positioning Service
Performance Standard, Washington DC. Available at
http://www.navcen.uscg.gov/gps/geninfo/ (September
2004)
[4] Abwerzger G., Lechner W. (2002): GPS and Loran-C
A raw data based integration method, presented at the
3
rd
International Symposium in Integration of Loran-
C/Eurofix and EGNOS/Galileo, 11-12 June 2002,
Munich, Germany.