Journal of Global Positioning Systems (2004)
Vol. 3, No. 1-2: 79-84
First results from Virtual Reference Station (VRS) and Precise Point
Positioning (PPP) GPS research at the Western Australian Centre for
Geodesy
N. Castlede n
Western Australian Cent re for Geodesy, Curtin University of Technology, GPO Box U1987, Perth WA 6845, Australia
e-mail: [e-mail]; Tel: [phone]; Fax: [phone]
G.R. Hu
Western Australian Cent re for Geodesy, Curtin University of Technology, GPO Box U1987, Perth WA 6845, Australia
e-mail: [e-mail]; Tel: [phone]; Fax: [phone]
D.A. Abbey
AAMHatch Pty Ltd, 23 Hamilto n Str ee t , Subiaco, WA 6008, Australia
e-mail: [e-mail]; Tel: [phone]; Fax: [phone]
D. Weihing
Geodetic Institute, University of Karlsruhe, Englerstr.7, D-76128 Karlsruhe, Germany
e-mail: [e-mail]; Tel: [phone]; Fax: [phone]
O. Øvstedal
Department of Mathematical Sciences and Technology , Agricultural University of No rway, P.O. Box 5003, N-1432 Ås, Norway
e-mail: [e-mail]; Tel: [phone]; Fax: [phone]
C.J. Earls
AAMHatch Pty Ltd, 23 Hamilto n Str ee t , Subiaco, WA 6008, Australia
e-mail: c.earls@aam hatch.com.au; Tel: [phone]; Fax: [phone]
W.E. Featherstone
Western Australian Cent re for Geodesy, Curtin University of Technology, GPO Box U1987, Perth WA 6845, Australia
e-mail: [e-mail]; Tel: [phone]; Fax: [phone]
Received: 15 Nov 2004 / Accepted: 3 Feb 2005
Abstract. Over the past 18 months, a team in the Western
Australian Centre for Geodesy at Curtin University of
Technology, Perth, has been researching the optimum
configurations to achieve long-range and precise GPS-
based aircraft positioning for subsequent airborne
mapping projects. Three parallel strategies have been
adopted to solve this problem: virtual reference stations
(VRS), precise point positioning (PPP), and multiple
reference stations (MRS). This paper briefly summarises
the concepts behind the PPP and VRS techniques,
describes the development and testing of in-house
software, and presents the latest experimental results of
our research. Current comparisons of the PPP and VRS
techniques with an independently well-controlled aircraft
trajectory and ground-based stations in Norway show that
each deliver precisions of around 3 cm. However, the
implementation of more sophisticated error modelling
strategies in the MRS approach is expected to better
deliver our project’s objectives.
Key words: GPS, kinematic positioning, VRS, PPP

80 Journal of Global Positioning Systems
1 Introduction
For many years, robust and reliable long-range kinematic
positioning has been the desire of airborne mapping
projects that utilise relative carrier-phase GPS as the
primary positioning technique (e.g., Cannon et al., 1992;
Colombo and Evans, 1998). However, this has been
difficult to achieve in a production environment because
of the de-correlation of, notably atmospheric, GPS-
related errors as the rover-to-basestation receiver distance
increases. This is somewhat problematic in Australia,
where airborne mapping projects are usually undertaken
in remote areas in which geodetic control is sparse and
ground access is difficult, thus significantly increasing
project costs.
There are currently three viable approaches to tackle this
problem:
1. The multi-basestation or multiple reference station
(MRS) approach (e.g., Han and Rizos, 1996; Wübbena
et al., 1996; Raquet, 1998; Wanninger, 1999, 2002;
Vollath et al., 2000; Fotopoulos and Cannon, 2001);
2. The virtual reference station (VRS) concept (e.g.,
Vollath et al., 2000; Higgins, 2002; Wanninger, 2002;
Hu et al., 2003); and
3. The precise point positioning (PPP) technique (e.g.,
Kouba and Heroux, 2001; Gao and Shen, 2001;
Witchayangkoon, 2000).
This paper describes initial results of our attempts to
implement the VRS and PPP approaches; our MRS
approach will be described in a later paper. Here, we
briefly describe the VRS and PPP concepts, followed by
results of some of our initial experiments using our in-
house-developed software. These show that both
techniques can deliver ~3 cm precision, but we expect
that improved error modelling from the MRS approach
may be better still.
2 The VRS concept
The VRS concept is a derivative of the MRS approach,
but differs in that the surrounding reference GPS
receivers are instead used to determine ‘synthetic’ dual-
frequency code and carrier-phase GPS data at a virtual
basestation that is located close to the user’s receiver.
The user’s and the virtual basestation GPS data are then
processed in single baseline mode to determine the
coordinates of the user’s receiver. This can be achieved
in near-real-time (akin to real-time kinematic, RTK) or
post-processed modes, but the near-real-time mode
requires high-bandwidth telecommunications among the
reference stations, master control station (where the
virtual GPS data are computed) and subsequently to the
user.
Of pragmatic benefit, the VRS approach does not require
an actual physical reference station close to the user, but
does require a MRS network of GPS stations surrou nding
the area of interest. The VRS concept allows the user to
access data from a virtual GPS reference station at any
location interpolated within the network coverage area.
Also, the VRS approach is more flexible in terms of
permitting users to use their current receivers and
software without involving any special software or
communications equipment (if used in post-processed
mode) to simultaneously manage data from all the
reference stations.
With VRS, users within the MRS network can operate at
distances greater than conventional RTK or fast/quick-
static GPS modes (typically 10 km and 20 km,
respectively) without degrading accuracy. Under ideal
conditions, the VRS approach can deliver single-point
coordinate accuracies of a few centimetres for a MRS
network of re f e rence stations sepa rat e d by 50-70 km .
With regard to our application-specific problem of
aircraft positioning, it should be noted that the VRS
concept was initially developed for fast/quick-static or
RTK users working in fairly small (less than 10 km)
areas. In such conditions, one approximate user position
is sufficient for the process of VRS data generation. For
long-range pure kinematic GPS positioning, however, the
rover may be moving over longer distances. Therefore, it
is necessary to frequently update the position of VRS so
that the VRS-to-rover separation is not too large, ideally
less than 10 km.
As such, we have modified the VRS method proposed by
Wanninger (2002) for long-range post-processed
kinematic GPS applications. In this modified method, the
VRS is referred to a fixed position according to the
rover’s initial approximate position, and the corrections
are applied to the rover’s trajectory. When the rover’s
current approximate position becomes greater than 10 km
from the initial VRS position, a new VRS is created.
This process continues each time the distance reaches 10
km, essentially making the VRS ‘follow’ the aircraft’s
trajectory.
3 The Curtin-AAMHatch VRS software
Post-processing V RS software for medium- to long -range
(30-100 km) kinematic positioning has been developed in
the Western Australian Centre for Geodesy at Curtin
University of Technology, Perth. The software generates
modified (as described above) VRS data for the pure
kinematic user in RINEX format so that the VRS-to-user
distance is always less than 10 km. Because of the
elevation difference between the ground-based and
airborne GPS receivers, the tropospheric delay must be
treated carefully.

Castleden et al.: First results of VRS and PPP GPS positioning research 81
Based upon our earlier tests, the UNB3 models (Collins
and Langley, 1996) with Niell mapping functions (Niell,
1996) are used in the in-house software to compute the a
priori troposphere delay. The post-processing mode of
operation allows for more careful treatment of the GPS
data and avoids the high-bandwidth telecommunication
problems commonly encountered in RTK applications.
For instance, we use the International GPS Service’s
(IGS) Final Product orbits instead of the broadcast GPS
ephemeri des .
4 VRS experiments and results
A dual-frequency airborne GPS dataset from 08:25 to
09:25 (GPS time) on 8 May 2002 was arbitrarily selected
from a test flight in Norway, with a data collection rate of
1 Hz [for further details about the whole flight period, see
Kjorsvik et al. (2004)]. This test used three Norwegian
reference stations (SAND, SORH and OE23; Fig 1) to
form a MRS network, also with the sampling rate of 1
Hz. The reference stations were equipped with dual-
frequency Trimble MS750 receivers and TRM41249.0
antennas. The distance between the reference stations
varies from 60-70 km. Figure 1 shows the test network
and the trajectory of the flight during the test period.
The VRS data were generated using our modified VRS
methods and software, then the aircraft trajectory
determined in a single-baseline mode with respect to each
‘following’ VRS station. The reference trajectory of the
aircraft is known to an accuracy of several centimetres
from comparisons with different software and
independently determined positions from an aerial
triangulation (Kjorsvik et al., 2004). The modified-VRS-
estimated positions were then compared with the
reference trajectory.
Fig 1. The one-hour airc raft flight trajectory (blue) and the three-
reference-station Norw eg i an n e tw o rk us ed fo r the
modified VRS tests (UTM projection)
Fig 2. Epoch-by- epoch differences (m et re s) between the modified
VRS results and the reference trajectory for the one-hour
airborne kinematic GPS test in Norway
STD Mean Max Min
North 0.031 -0.005 0.068 -0.076
East 0.031 0.010 0.097 -0.107
Height 0.078 -0.035 0.157 -0.219
Table 1. Descriptive statistics of the differences (metres) between
the modified VRS results and the reference trajectory for
the one-hour airborne kinematic GPS test in Norway
Figure 2 shows the position differences in the easting,
northing and height components, and Table 1 summarizes
the results of the comparisons conducted during the one-
hour test period. From Table 1, the standard deviations
of north, east and height components are 31 mm, 31 mm
and 78 mm, respectively, when using the modified VRS
method. Importantly, this is commensurate with the
expected accuracy of the reference trajectory.
As expected, the precision of the height component is a
factor of 2 to 3 times worse than for the horizontal
components (Table 1). Since the accuracy of GPS
positioning is inherently worse in the height component,
it is necessary to pay careful attention to the a priori
tropospheric delay, which is the subject of our current
work (e.g., Hu et al., 2005). The oscillations of the
horizontal coordinate precision in Fig 2 are probably
caused mainly by multipath from the aircraft. The main
sources for the vertical oscilla tions are most prob ably due
to the combination of multip ath and residual tropospheric
errors.
5 The PPP concept
Classical GPS point positioning, also known as
standalone positioning, involves only one code-tracking
GPS receiver. That is, one GPS receiver simultaneously
tracks the code pseudoranges from four or more GPS
satellites to determine its position (e.g., Hofmann-
Wellenhof et al., 2001). After the discontinuation of

82 Journal of Global Positioning Systems
Selective Availability, the expected horizontal
positioning accuracy of this classical point positioning
approach is about 22m (2DRMS) or better (e.g., Shaw et
al., 2000; Witchayangkoon, 2000).
However, it is also possible to determine single point
positions from both code and carrier-phase data. Using
precise GPS orbit and clock products (i.e., time
corrections to the satellite clo cks) available fro m the IGS,
it has been shown that carrier-phase-based single point
positioning can be improved to decimetre accuracy lev els
(e.g., Witchayangkoon, 2000; Gao and Shen, 2001;
Kouba and Heroux, 2001). This is through the use of
ionosphere-free, undifferenced code pseudorange and
carrier-phase measurements, together with the precise
IGS orbit and clock products. This approach is
commonly known as Precise Point Positioning (PPP).
The PPP approach is very similar to single-point code
pseudorange positioning achieved from a hand-held GPS
receiver, except that it adds the more precise carrier-
phase GPS data, as well as improved orbits and satellite
clock corrections. Under ideal conditions, the PPP
concept can deliver single point accuracies of 5-10 cm
irrespective of baseline length.
PPP is relatively new and is now attracting much
attention internationally (e.g., Bisnarth et al., 2002).
Importantly, it requires only one dual-frequency carrier-
phase GPS receiver and thus avoids the expense and
logistics of deploying a network of GPS receivers
surrounding the area of interest, as is needed for the MRS
and VRS techniques. However, it uses a (global) MRS
network by ‘stealth’ through the data provided by the
IGS.
6 The Curtin-AAMHatch PPP software
We are developing in-house software completely from
scratch to implement the PPP technique, in post-
processing mode, according to two mathematical models.
• PPP Strategy 1 takes the model of Kouba and Heroux
(2001), which uses ionosphere-free linear combination s
of L1 and L2 carrier-phase data, but cannot deliver
integer-ambiguity-fixed solutions (i.e., it only delivers
float solutions).
• PPP Strategy 2 uses the model of Gao and Shen
(2001), which uses a ionosphere-free linear
combination of L1 and L2 carrier-phase data that
reduces code pseudorange noise and tries to return the
integer properties of the ambiguities.
Currently, only PPP Strategy 1 is coded in our software;
Strategy 2 is under develo pm ent .
Since the PPP technique does not enjoy the cancellation
of errors found in relative/differential GPS techniques,
several corrections have been included in our software.
In addition to the IGS Final products described earlier,
these are: relativity corrections for the satellites’ orbital
eccentricity and the Earth’s rotation (the Sagnac
correction), satellite antenna offsets from their centre of
gravity (notably the so-called Y-bias), tropospheric
delays, phase windup (essentially the carrier-phase at the
satellite) and solid-Earth tides. A sequential filter and
smoothing procedure has also been implemented in our
PPP software, where the ambiguities are essentially
carried forward from epoch to epoch.
7 PPP experiments and results
While we seek airborne kinematic positioning, we have
undertaken our PPP Strategy 1 software testing with
ground-based static GPS data, but processed in kinematic
mode. Here, the ‘error’ values are the epoch-by-epoch
differences between the PPP-estimated position and the
well-controlled ground-based station coordinates. While
we have compiled a large number of well-coordinated
static GPS datasets, the few results presented below for
the test data in Norway are typical of those achieved with
our other datasets. The use of an epoch-by-epoch PPP
solution partially simulates a (ground-based) kinematic
example.
Five GPS stations (HONE, ARNE, KONG, SORH,
SAND; Fig 3) from a MRS network in Norway were
used. Although the distances between these five stations
range between 63 km and 132 km, this does not matter
because the PPP method is baseline-length-independent.
Each station was equipped with Trimble MS750 dual-
frequency receivers and TRM41249.0 antennas. The data
set used to test the PPP software was measured between
07:45 and 10:15 (GPS time) on 8 May 2002 with a
sampling rate of 1 Hz.
The 3D coordinates of these five reference stations are
known to ~1 cm in the horizontal and ~3 cm in the
vertical on the EUREF89 datum. These were determined
from processing three days of static GPS data with the
GIPSY scientific software. The north, east and up
component ‘error’ values, epoch-by-epoch, were
computed by subtracting the PPP-estimated position from
the reference coordinates. One graphical example is
presented in Fi g 4.

Castleden et al.: First results of VRS and PPP GPS positioning research 83
Fig 3. The five-re fere n c e -station Norwegian network
used for the PPP tests (0.1° ~ 11 km)
Fig 4. Epoch-by-epoch differences (metres) between the PPP results
and the known coordina t e s o f the static station KONG for t h e
three-hour (sim ulated) kinematic GPS t est in Norway
STD Mean Station
North East Height North East Height
KONG 0.028 0.048 0.074 0.036 -0.170 0.157
HONE 0.030 0.032 0.063 0.042 -0.198 0.034
ARNE 0.033 0.030 0.053 0.038 -0.128 -0.016
SORH 0.023 0.033 0.035 0.062 0.006 0.118
SAND 0.005 0.005 0.002 -0.008 -0.034 0.116
Table 2. Descriptive statistics of the differences (metres) between the
PPP results and the known coordinates of the static stations for the
three-hour (sim ulated) kinematic GPS t ests in Norway
Table 2 summarises the PPP ‘error’ values for all five
stations. It can be seen from Table 2 that the PPP
Strategy 1 solutions achieve around 30 mm precisions for
the north, east and height components in terms of
standard deviations, but this depends on the GPS data
quality from the stations. Curiou sly, the height errors are
less than expected in some cases.
8 Concluding remarks
We have presented some of our preliminary results for
long-range kinematic GPS positioning using two post-
processing software packages developed in-house at the
Western Australian Centre for Geodesy. For the
modified VRS software, test results using one-hour of 1
Hz real airborne kinematic GPS data in Norway show ~3
cm precision for the horizontal components and ~8 cm
precision for the height component. For the PPP
software, epoch-by-epoch static (i.e., simulated ground-
based kinematic) test results using three hours of 1 Hz
data in Norway indicate ~3 cm precision in all three
coordinate components. While these results are
extremely encouraging, and commensurate with studies
conducted by other authors, we are now focussing on the
post-processed MRS approach (e.g., Hu et al., 2005),
which will allow the use of network constraints and
improved (particularly atmospheric) error modelling.
Acknowledgements:
The Australian Research Council and AAMHatch Pty Ltd
jointly fund this research
References
Bisnarth SN, Beran T, Lngley RB (2002) Precise platform
positioning with a single GPS receiver, GPS World 13(4):
42-49.
Cannon ME, Schwarz KP, Wei M (1992) A consistency test of
airborne GPS using multiple monitor stations, Bulletin
Géodésique 66(1): 2-11
Collins P, Langley R (1996) Limiting factors in tropospheric
propagation delay error modelling for GPS airborne
navigation. Proceedings of The Institute of Navigation
52nd Annual Meeting, 19–21 June, Cambridge,
Massachusetts, USA, pp 519-528
Colombo OL, Evans AG (1998) Testing decimeter-level,
kinematic, differential GPS over great distances at sea
and on land, Proceedings of ION GPS 1998, 15-18 Sept,
Nashville, Tennessee, USA, pp 1257-1264
Fotopoulos G, Cannon ME (2001) An overview of multi-
reference station methods for cm-level positioning, GPS
Solutions 4(3): 1-10
Gao Y, Shen X (2001) Improving ambiguity convergence in
carrier phase-based precise point positioning.
Proceedings of ION GPS 2001, 11-14 September, Salt
Lake City, Utah, USA, pp 1532-1539

84 Journal of Global Positioning Systems
Han S, Rizos C (1996) GPS network design and error
mitigation for real-time continuous array monitoring
system, Proceedings of ION GPS-1996, Sept 17-20,
Kansas City, Missouri, USA, pp 1827-1836
Higgins MB (2002) Australia’s changing surveying
infrastructure from marks in the ground to virtual
reference stations. Proceedings of FIG XXII International
Congress, April 19-26, Washington, USA, CD-ROM paper
TS5.6
Hofmann-Wellenhof B, Lichtenegger H, Collins J (2001)
Global positioning system: theory and practice. 5th edn.
Springer, Vienna New York
Hu GR, Khoo HS, Goh PC, Law CL (2003) Development and
assessment of virtual reference stations for RTK
positioning. Journal of Geodesy 77(5-6): 292-302
Hu G, Abbey DA, Castleden N, Featherstone WE, Earls CJ,
Ovstedal O, Weihing D (2005) An approach for
instantaneous ambiguity resolution for medium- to long-
range multiple reference station networks. GPS Solutions
9(1):1-11
Kjorsvik N, Ovstedal O, Svendsen JGG (2004) Long range
differential GNSS for positioning of airborne sensors. in:
Sansò, F. (ed) A Window on the Future of Geodesy,
Springer, Berlin Heidelberg New York (in press)
Kouba J, Héroux P (2001) Precise Point Positioning using IGS
orbit and clock products. GPS Solutions 5(2): 12-28
Niell AE (1996) Global mapping functions for the atmosphere
delay at radio wavelengths. Journal of Geophysical
Research 101(B2): 3227-3246
Raquet J (1998) Development of a method for kinematic GPS
carrier phase ambiguity resolution using multiple
reference receivers, UCGE report 20116, Dept of
Geomatics, Univ of Calgary
(http://www.geomatics.ucalgary.ca/gradtheses.html).
Shaw M, Sandhoo K, Turner D (2000) Modernization of the
Global Positioning System. GPS World 11(9): 36-44
Vollath U, Buecherl A, Landau H, Pagels C, Wagner B (2000)
Multi-base RTK positioning using virtual reference
stations. Proceedings of ION GPS-2000, 19-22 Sept, Salt
Lake City, Utah, USA, pp 123-131
Wanninger L (1999) The performance of virtual reference
stations in active geodetic GPS-networks under solar
maximum conditions. Proceedings of ION GPS-1999,
Nashville, Tennessee, USA, pp 1419-1427
Wanninger L (2002) Virtual reference stations for centimeter-
level kinematic positioning, Proceedings of ION GPS-
2002, 24-27 Sept, Portland, Oregon, USA, pp 1400-1407
Witchayangkoon B (2000). Elements of GPS Precise Point
Positioning. Ph.D. dissertation, Dept of Spatial
Information Science and Engineering, Univ of Maine,
USA, 265 pp
Wübbena G, Bagge A, Seeber G, Böder V, Hankemeier P
(1996) Reducing distance dependent errors for real-time
precise DGPS applications by establishing reference
station networks. Proceedings of ION GPS-1996, Sept 17-
20, Kansas City, Missouri, USA, pp 1845-1852