Journal of Global Positioning Systems (2003)
Vol.2, No.2:139-143
Network Differential GPS: Kinematic Positioning with NASA’s
Internet-based Global Differential GPS
M. O. Kechine, C.C.J.M.Tiberius, H. van der Marel
Delft InstituteofEarth ObservationandSpace Systems,DelftUniversityof Technology,Kluyverweg1,2629 HSDelft,
The Netherlands
Received:12 November2003/ Accepted: 22 December2003
Abstract. Recent developments in precise GPS position-
ing have concentrated on the enhancement of the GPS Net-
workarchitecturetowards the processingofdata from per-
manent reference stations in real-time, and the extension of
the DGPS service area to the continental and global scale.
The latest Global Differential GPS, as introduced by JPL,
allows forseamlesspositioning available acrosstheworld.
This contribution presents the results of an independent ex-
perimental verification of decimeter kinematic positioning
accuracy with NASA’s Global DGPS system.This veri-
fication was carried out in the Netherlands, by means of
both a static and a kinematic test.The standard deviations
of individual real-timepositions were about10 cm forthe
horizontal components and about 20 cm for the vertical
component.The latency ofthe global corrective informa-
tion in the kinematic test was generally 7 to 8 seconds and
more than 99% of the global corrections were available
with thenominal 1-secondinterval.
These results confirm that single receiver kinematic posi-
tioning with decimeteraccuracy is achievable by using fa-
cilities provided by the GDGPS system.
Key words:Network Differential GPS, IGDG, kinematic
positioning, real-time dm-accuracy
1 Introduction
1.1Recent trends and developments inprecise
positioning
Relative positioning with GPS and Differential GPS
(DGPS) both involve the positioning of asecond receiver
with respect to a reference station.As both stations sim-
ilarly experience—dependingontheirinter-distance—
the effects of satellite orbits/clocks and atmospheric de-
lays, the relative position islargely insensitivetomismod-
elling of these effects and their errors.
The concepts of relative positioning with GPS and Dif-
ferential GPS haveexisted for sometwentyyears.Until
recently,thesetwo fieldshave developed relatively inde-
pendently from each other.Two new trends inboth DGPS
positioning and GPSReal-Time Kinematic(RTK) survey-
ing includemoving fromscalar corrections(from oneref-
erence station) to (state) vector-’corrections’,basedona
network of reference stations;and the processing of the
data,also for the global high precision IGS-type (Inter-
national GPS Service) of applications, is movingtowards
real-time execution.As a resultthe traditional distinction
between preciserelative positioningwith GPSand DGPS
diminishes;instead, one consistent family of applications
emerges, sharing a common concept andcommon algo-
rithms, that could betermed Network-based Differential
GPS (NDG).
1.2 Network
Initially, systems for DGPS started with one reference sta-
tion, and one or more mobile receivers (rovers) ina local
area.Later, the service area of Differential GPS was ex-
tended from local toregional and national, andeventually

140 Journal of Global Positioning Systems
to the continental scale with Wide Area DGPS (WADGPS)
systems such as WAAS (Wide Area Augmentation Sys-
tem) in the US and EGNOS (EuropeanGeostationary Nav-
igation Overlay Service) in Europe.Logically, the last step
is GlobalDGPS,as introduced byJPL(M¨ ullersch¨ on etal.,
2001a).Thus making seamless DGPS positioning avail-
able across the world.Theadvantage is thatcostly infras-
tructure is nolonger needed, however, theuser has torely
on theUS Department ofDefence (DoD)for GPSdata,
on a globalinfrastructureofactive GPS reference stations,
and on NASA’sJPLfor thecorrective information.
1.3Real-time products
The Internet-basedGlobalDifferentialGPS(IGDG) sys-
tem aimsatreal-time precise positiondeterminationof a
single receiver either stationary or mobile, anywhere and
anytime.The concept of Precise Point Positioning (PPP)
was introduced in theearly1970s, for more details re-
fer tothekeyarticlebyZumberge et al. (1997).Precise
Point Positioning utilizes fixed precise satelliteclock and
orbit solutions for single receiverpositioning.Thisis a key
to stand-alone precise geodetic point positioningwith cm
level precision.
Over the pastseveral years the quality of theRapidIGS
satellite clock and orbit products has improved to the cm
level.Today the IGS Rapid service provides the satellite
clock/orbit solutions within one day, with almost the same
precision astheprecise finalIGS solutions(IGS, 2004).
A good agreement between satellite clock errorestimates
produced by 7 Analysis Centers (AC) contributing to the
IGS is reached.Theseestimatesagree within 0.1–0.2ns
or 3–6cm.CurrentlyIGSorbitswithafew decimeter
precision, canbemade availablein(near) real-time.Ultra-
rapid/predicted ephemerides are available twice each day
(at 03:00and 15:00 UT),and cover 48 hours.The first
27 hoursare based onobservations, the second partgives
a predictedorbit.Itallowsone toobtain highprecision
positioning resultsin thefieldusingtheIGS products.
1.4Dissemination of corrective information
Traditionally,DGPS-corrections are broadcastover a
radio-link from reference receiver to rover.With IGDG,
corrections aredisseminatedovertheopenInternet.The
user can access the very modest correction data streamus-
ing a (direct and) permanent network connection, or over
the public switched telephone network (PSTN), possibly
using an Asynchrone Digital Subscriber Line (ADSL). For
a movinguser accessis possibleusing mobile (data)com-
munication by cellular phone (possiblyGeneral Packet Ra-
dio Service(GPRS) or theUniversalMobile Telecommu-
nication System (UMTS) in future) or satellite phone.For
commercial use threeInmarsat geosynchronous commu-
nication satellitesare utilizedto relay thecorrection mes-
sages on their L-band global beams.Thethree satellites (at
100◦W(Americas), 25◦E(Africa), 100◦E(Asia Pacific))
provideglobal coveragefromlatitude−75◦to +75◦.
2Internet-BasedGlobal Differential GPS
In Spring2001, the JetPropulsionLaboratory(JPL)of
the National Aeronautics Space Administration (NASA)
launched Internet-based GlobalDifferential GPS (IGDG).
Compared withtraditionalDifferentialGPS(DGPS)ser-
vices, the position accuracy improves by almostone order
of magnitude.An accuracy of10cmhorizontal and 20 cm
vertical is claimed for kinematic applications, anywhere
on theglobe, andatanytime.Thislevel ofpositionac-
curacyisvery promising forprecisenavigation ofvehicles
on land, sea vessels and aircraft, and for Geographic In-
formation System(GIS) data collection, forinstance with
construction worksandmaintenance.
A subset of some 40 reference stations of NASA’s Global
GPS Network (GGN) allows forreal-timestreaming of
data to a processing center,thatdetermines and subse-
quently disseminates over the open Internet, in real-time,
precise satellite orbits and clocks errors, as global differen-
tial corrections totheGPSbroadcast ephemerides (as con-
tained in the GPS navigation message).Anintroduction
to IGDG can be found in M¨ ullersch¨ on et al. (2001a) and
on IGDG(2004).Technical details are given inBar-Sever
et al. (2001) and M¨ ullersch¨ on et al. (2001b).
Internet-based users can simply download the low-
bandwidth correction data stream into a computer, where
it willbe combinedwith raw datafrom theuser’sGPS re-
ceiver.The user’s GPS receiver mustbe adual frequency
engine and beof geodeticqualityin ordertoextract maxi-
mum benefit fromtheaccuratecorrections.
The final, but critical element in providing an end-to-end
positioning and orbit determination capability,is the user’s
navigation software.In order to deliver 10 cm real-time
positioning accuracy the software must employ the most
accurate models for the user’s dynamics and the GPS mea-
surements.Forterrestrialapplicationsthesemodels in-
clude tropospheric mapping function, Earthtides,periodic
relativity effect, and phase wind-up, see also the review
in Kouba and H´ eroux (2001).In addition to these mod-
els, the end-user version ofthe Real-TimeGipsy (RTG)
softwareemploys powerfulestimationtechniques for opti-
mal positioning or orbit determination, including stochas-
tic modelling, estimation of tropospheric delay, continuous
phase smoothing and reduced dynamics estimation with
stochastic attributes foreveryparameter.
Results ofstatic post-processing precisepoint positioning
are shownin, for instance,the articles Koubaand H´ eroux

Kechine et al:Network DGPS: Kinematic Positioning with IGDG141
9h10m10h 10h30m 11h 11h30m 12h
−3
−2
−1
0
1
2
3differences for rover antenna (Ashtech). April 3rd, 9h10m − 11h47m
time [min]
[m]
North
East
Height
Fig. 1 Coordinate time seriesfor the receiveronboard theboat in the
kinematic test; differences with ground-truth trajectory:wet troposphere
is estimated as a constant (strategy A).
(2001) andGao andShen (2002).Furthermore, kinematic
post-processing point positioning results can be found e.g.
in Bisnathand Langley (2002).
3Kinematic positioning withIGDG
3.1 Results
An independent experimental verification of the IGDG
system has been carriedout, by meansof botha staticand
kinematic testin theNetherlands.The GPSdata collected
during five consecutive days (static test) and three hours
(kinematic test) were processed using the filter algorithm
implemented intheGIPSY-OASIS IIsoftware, seeGrego-
rius (1996)and Gipsy(2004).
In the static test,the means of the position coordinates,
taken over individual days of data, agree with the known
reference at the 1–2cm level.The IGDG position solu-
tions appeared to be free of systematic biases.The stan-
dard deviations of individual real-time position solutions
were 10 cm for thehorizontal components and 20 cm for
the vertical component.The position coordinate estimators
were correlated overabouta1hour timespan.
In the kinematic test, which was carried out with a small
boat, the means of thecoordinate differences withan ac-
curate ground-truthtrajectoryover thealmost3hourpe-
riod were at the 1–2dm level.The standard deviations
of individual positions weresimilar to values foundin the
static test, 10 cm for the horizontal components, and 20 cm
for the verticalcomponent.Morethan 99% of the IGDG-
corrections were received with the nominal interval of 1
second,inthe field via mobile communication using a
9h10m10h 10h30m 11h 11h30m 12h
−3
−2
−1
0
1
2
3differences for rover antenna (Ashtech). April 3rd, 9h10m − 11h47m
time [min]
[m]
North
East
Height
Fig. 2 Coordinate time seriesfor the receiveronboard theboat in the
kinematic test; differences withground-truthtrajectory: both wet tropo-
sphere and troposphere gradients are esimated stochastically (strategy B).
GPRS cellular phone.The latency of the corrections was
generally 7to 8seconds, for more details see Kechine et al.
(2003).
The resultspresented inthis contribution donot relyon
the Internet corrections,but on thereal-time JPLorbitand
clock solutions instead(RTG, 2004), which are stated to
be 100% consistent(Bar-Sever,2003).
Figure 1 shows differences of the filtered position esti-
mates for an Ashtech receiver on theboat used for the kine-
matic test, with a cm-level ground-truth trajectory.For this
case, the wet troposphere (zenith delay) wasestimated as
a constant parameterfor thewholetime span(strategy A).
The kinematic test resultsinfigure 2 represent a strategy
with both the wet troposphere and troposphere gradients
estimated stochastically (strategyB).For both strategies,
the initial value for the dry zenith tropospheric path delay
was computed by GIPSY (a-priori model), whereas the ini-
tial value for the wet part was set to 10 cm by default.The
boat coordinates were modelled as white noise; the process
noise was 100min order toaccommodatefor dynamics of
the boat andavoid possible divergenceproblems.
A comparison of these resultsallowsone to conclude that
estimation of troposphere zenith delays and gradients (as
stochastic processes) in the case ofsingle receiver precise
kinematic positioning, mightsignificantly affect filterini-
tialization and render the filtered estimates vulnerable to
various error sourcescapableofdegrading the positional
accuracy.For instance, as additional analysesshowed, a
peak inthe Heightbetween9:40 and9:50 infigure 2is
most likely causedby adeviating clockerror estimate for
one ofthe satellitesin theJPL real-timeephemerides at
epoch 9:45.At the same time, the peak is present in fig-

142 Journal of Global Positioning Systems
Table 1 Mean of position differences, in kinematic test; filter initialization
is left out.
North(cm) East(cm) Height(cm)
strategyA−5.9 15.5−13.1
strategyB−2.2 18.9−24.7
Table2 Standard deviation of position differences, in kinematic test; filter
initialization is left out.
North(cm) East(cm) Height(cm)
strategyA6.2 14.2 15.8
strategyB8.0 12.3 20.3
ure 1, but the magnitude of the corresponding Height com-
ponent deviation is noticeablydecreased.Becausethetro-
posphere gradientsaregenerallysmaller than 1cm, they
have aminor impacton kinematicpositioning results,and
their estimationseemsnottobe necessary inthecaseof
kinematic positioning at the dm level.Due to quiet tro-
pospheric circumstances duringthe kinematic test thewet
troposphere delaycould also beleft outin this case(strat-
egyA).
In ordertodemonstrate how thehorizontalcomponents
convergence profile is influenced by lessaccurate or er-
roneous initialposition estimates,theinitial values for
the North and East positioncomponents wereartificially
shifted by10 m,as maybe thecase foran approximately
known initial horizontal position obtained from a stan-
dalone GPSsolution forexample.Analysisof theerro-
neous initial position results showed that the behaviour of
the horizontal position component during the filter initial-
ization in case of strategyA remained noticeably stable.
The corresponding boat positioning results were nearly
identical tothosepresented infigure1. In thecaseofstrat-
egyBthe largeinitialdeviationsreducedin a fewminutes.
The mean and standard deviation of the position differ-
ences in the kinematic test at a 1 secondinterval are given
in tables 1 and 2.It is to be noted that theperiod with-
out thefilter initializationis considered here.The first40
minutes were not included forstrategyB and the first 20
minutes were notincludedforstrategy A.
3.2 Analysis
Additional tests were performed in order to obtain a bet-
ter understanding ofthekinematic positioningcapabilities
with IGDG,andto assessthe impactof someimportant
factors (filter convergence, GPS orbit products quality,etc)
on real-time kinematic positioning.Only strategy Bis con-
9h10m10h 10h30m 11h 11h30m 12h
−1.5
−1
−0.5
0
0.5
1.0
1.5 Differences for marker #22. April 3rd, 9h10m − 11h47m
time [min]
[m]
North
East
Height
Fig. 3 Coordinatetimeseriesforthe(stationary)referencestationduring
the kinematic test; differences with the ground-truthposition(strategy B).
sidered for the kinematic test computations.
Figure 3 demonstrates the position estimates as differences
with theground-truth position, for the(nearby) stationary
reference receiver installed on a well-surveyed reference
markerin Delft.Dmlevel accuracy isevident throughout
the test period.Note the difference inscale of the vertical
axis withthe precedinggraphs.
The kinematic processingprocedure wasrepeated witha 5-
min sampling interval in order to avoid interpolation of the
JPL’sReal-TimeGPS satelliteorbits/clocks (RTG, 2004).
The positioning results for this case can be seen in figure 4.
One cannote that thetime series is relatively smooth and
without any significant variability.Thestandard deviations
were about 5cm for thehorizontal components and 9cm
for the vertical component in case of the Real-TimeGPS
satellite orbits/clocks, andabout 3cmfor thehorizontal
components and5cm for the vertical component in case
of theJPL’s Final GPS satelliteorbits/clocks.
4Further research
A number of additional testsare to be carried out to pro-
vide a betterinsight into the filter initializationproblem
in case ofprecise real-timekinematic positioningof asin-
gle receiver.The task is to seek fast and smooth conver-
gence ofthefilteredposition estimates during thefirstsec-
onds after the filtering process starttime.A primary in-
terest would be to establish whether the constrained tro-
posphere errors (taken from a-priorimodels)arecapable
of decreasingthe filter convergence time.This problem
can be importantfor regionswith a high concentrationof
water vapour inthe atmosphere and largewet delay vari-
ations (e.g.Pacific region).It is tobe noted here that the

Kechine et al:Network DGPS: Kinematic Positioning with IGDG143
9h10m10h 10h30m 11h 11h30m 12h
−3
−2
−1
0
1
2
3differences for rover antenna. April 3rd, 9h10m − 11h47m
time [min]
[m]
North
East
Height
Fig. 4 Coordinate time seriesfor the receiveronboard theboat in the
kinematic testata5-min sampling interval;differences withground-truth
trajectory (strategy B).
kinematic test in this contribution was carried out in the
Netherlands with rathermoderate troposphere conditions.
More GPS data shouldbe processed in orderto assess the
repeatability of kinematicpositioning results with IGDG,
e.g.fordifferentseasons andweather conditions.Con-
versely, the Precise Point Positioning approach is apoten-
tial powerful technique to obtain accuratewetzenith tro-
pospheric path delayestimatesusingasingle receiver.
The GPS data processing strategyadoptedfor the kine-
matic test computations requires further refinement in or-
der to expand it tothe case of a receiver with high platform
dynamics (a receiver installed on a moving car, airborne
and spaceborne receivers).This will allow for acom-
prehensive analysis ofthe IGDG performance for aircraft
landings and takeoffs, and spacekinematicapplications.
The problem of single-receiver carrier phase ambiguity
resolution is one of the most importantand interesting
challenges to beinvestigated in thefuture,andthebenefits
of fixing integer ambiguities to theperformance of carrier
phase preciseGDGPS navigation requirefurther evalua-
tion.
References
Bar-Sever,Y.(2003). Personal communication. Jet Propalsion
Laboratory,Pasadena, CA.Summer2003.
Bar-Sever,Y., M¨ ullersch¨ on, R., Reichert, A., Vozoff, M., and
Young, L.(2001).NASA’s Internet-Based Global Differ-
ential GPS System.In Proceedings of NaviTec.ESA/Estec
- Noordwijk, The Netherlands.December 10–12, pp.65–
72.
Bisnath, S. andLangley, R. (2002).High-precision,kinematic
positioning withasingle GPS receiver.Navigation,49(3),
pp. 161–169.
Gao,Y. and Shen, X. (2002).ANew Method forCarrier-
Phase-Based Precise Point Positioning.Navigation, 49(2),
pp. 109–116.
Gipsy (2004).Gipsy-Oasis II software package.Internet URL:
http://gipsy.jpl.nasa.gov/orms/goa/.
Gregorius, T. (1996).Gipsy-OasisII: How it works...Depart-
ment of Geomatics, University of Newcastle upon Tyne.
IGDG (2004).Internet-basedGlobal Differential GPS. Inter-
net URL:http://gipsy.jpl.nasa.gov/igdg/.
IGS (2004).IGS Product Availability.Internet URL:
http://igscb.jpl.nasa.gov/components/prods cb.html.
Kechine, M. O., Tiberius, C., and van der Marel, H. (2003).Ex-
perimental verification ofInternet-based Global Differen-
tial GPS. In Proceedings of IONGPS/GNSS2003, Port-
land, OR,September 9–12, pp. 28–37.
Kouba, J. and H´ eroux, P. (2001).Precise Point Positioning Us-
ing IGSOrbit and Clock Products.GPS Solutions, 5(2),
pp. 12–28.
M¨ ullersch¨ on, R., Bar-Sever, Y., Bertiger, W., andStowers, D.
(2001a). NASA’sGlobalDGPS forHigh-PrecisionUsers.
GPS World,12(1),pp.14–20.
M¨ ullersch¨ on, R., Reichert,A.,Kuang, D., Heflin,M.,Bertiger,
W., andBar-Sever,Y.(2001b).Orbit Determinationwith
NASA’s HighAccuracy Real-TimeGlobal Differential
GPS System.In Proceedings of ION GPS-2001, Salt Lake
City,Utah, September 11–14, pp.2294–2303.
RTG(2004).JPL Real - Time GPS products.ftp server:
ftp://sideshow.jpl.nasa.gov/pub/15min/.
Zumberge, J.,Heflin, M., Jefferson, D., Watkins, M., and Webb,
F. (1997).Precise point positioning for the efficient and
robustanalysisofGPS data from large networks.Journal
of Geophysical Research, 102(B3), pp.5005–5017.