gage using gpstk gnss learning research seagal hanoi 2009

1SEAGAL Project Workshop. Hanoi, 2009.

Salazar, D., Hernandez-Pajares, M., Juan, J.M., Sanz, J.

Using the GPSTkfor GNSS learning and

research

Salazar D., Hernandez-Pajares M.,

Juan J.M. and Sanz J.

gAGE/UPC, Barcelona, Spain

Contents

• GPSTk overview

• Some current features

• GPSTk advantages and disadvantages

• How to get the GPSTk

• GPSTk basics

• GNSS Data Structures (GDS)– Processing paradigm

• Some processing examples

• Comparison with other GPS software

• Conclusions

GPSTk overview

• The GPSTk is:

– A library to write GNSS software– Includes example applications.

• The GPSTk is Free Software (LGPL):

– Both non-commercial and– commercial applications.

GPSTk overview

The LGPL free software license means that:

• The original code belonged to ARL:UT, but it was released to the public.

• New features added to the library are property of their authors, but them must be also released as LGPL.

• New software developed taking advantage of the GPSTk library (linking) is property of their authors.

GPSTk overview

The LGPL free software license means that:

• In summary, with the GPSTk you can develop both free and non-free (commercial) software.

• This license is very convenient both for companies and educational/research institutions.

GPSTk overview

• Initiated at ARL:UT, it now has several official developers around the world.

• The GPSTk is easy to extend and maintain, lowering programming costs.

• It is very well documented.

• Website:

http://www.gpstk.org

GPSTk overview

• It is written in ISO-standard C++.

• Design is object-oriented, easing maintenance and extensibility.

• It is VERY portable:

– Operative System portability:

• Works in Windows, Linux, Mac OSX, AIX, etc.

– Hardware portability:

• Big and small systems, both 32 and 64 bits.

GPSTk portabilityGUMSTIXMiniature computer

NOKIA Internet Tablets

Some current features

• Time conversions.

• RINEX files reading/writing:– Observation– Ephemeris– Meteorological

• Ephemeris computation:– Broadcast– SP3.

• Mathematical tools: Matrices, vectors, interpolation, numeric integration, LMS, W-LMS, Kalman filter, etc.

• Application development support:– Exceptions handling.– Command line framework.– Configuration files management.

• Code positioning, RAIM, DGPS and Precise Point Positioning (PPP) support.

• Several tropospheric models: Saastamoinen, Goad-Goodman, New Brunswick, Niell, etc.

• Classes for precise modeling: Phase

centers, Wind-up, gravitational delay,

• Tidal models: Solid tides, Ocean

loading, Pole tides

• Full support for Antex (antenna) files.

• Advanced GNSS Data Structures (GDS)

for data processing and management.

• Other features like:

– Stocastic models

– Probabilistic functions (normal, chi-square,

– Run-time programmable solvers!!!

GPSTk advantages

• Programmers don't have to reinvent the wheel:– Don't use your time to program, test and

debug common, non-interesting routines (like RINEX parsing, for instance)

– Use your time to learn and experiment. – Use your time to develop new techniques.

• You can trust on the GPSTk:– Its performance has been validated with

other state-of-the-art GPS software suites.

GPSTk advantages

• It includes data management facilities that support clean and easy to maintain source code.

• It is open-source, and therefore:

– Great for learning how complex algorithms work.

– It eases reimplementation in other languagues and/or hardware.

GPSTk disadvantages

• The programmer needs C++ knowledge to use it effectively.

• C++ is very flexible, but it may become complex and “exotic”.

• The compilation process is slower and more complex than using interpreted languages.

• Many engineers are more confortable with tools like Mathlab (although they are very slow).

How to get the GPSTk

• The GPSTk requires:– C++ compiler (g++, MS Visual C++, etc)– Jam or make– Optional: doxygen, perl, gnuplot

• Easy installation:– Download source code from GPSTk website or

subversion repository– Decompress if needed– Compile: “jam” is currently the easiest way– Generate documentation with “doxygen”

GPSTk basics: Vectors Vector<double> vect1(4);Vector<double> vect2(4);

for( int i = 0; i < 4; ++i ){ vect1(i) = 3.0 + i; vect2(i) = 5.0 2.5*i;}

Vector<double> vect3 = vect1 + vect2;

cout << vect1 << endl;cout << vect2 << endl;cout << vect3 << endl;

GPSTk basics: Matrices Matrix<double> A(3, 3, 5.0);Matrix<double> I = ident<double>(3);Matrix<double> B = A 2.0*I;Matrix<double> C(3, 3);

for( int row = 0; row < 3; ++row ){ for( int col = 0; col < 3; ++col ) { C(row, col) = row+0.9*B(row,col); }}

Matrix<double> CT = transpose( C );Matrix<double> D = inverseChol( CT*C );

GPSTk basics: Solvers Vector<double> vect(3, 0.0);vect(0) = 94.50;vect(1) = 112.01;vect(2) = 121.86;

Matrix<double> mat(3, 2, 0.0);for( int row = 0; row < 3; ++row ) for( int col = 0; col < 2; ++col ) mat(row, col) = 1.5 + row + 0.9*row*col;

SolverLMS solver;

solver.Compute( vect, mat );

cout << solver.solution(0) << " : " << solver.solution(1) << endl;

GPSTk basics: Time

DayTime epoch1(2009,10,28,10,30,0.0);

DayTime epoch2 = epoch1 + 3600.0;

cout << epoch1 << endl << epoch2 << endl;

cout << epoch1.GPSfullweek() << " " << epoch1.DOY() << " " << epoch1.GPSsecond() << endl;

GPSTk basics: Position

Position posBCN( 4789031.0, 176583.0, 4195015.0 );

Position posHANOI;posHANOI.setGeodetic( 21.033, 105.85, 308.0 );

Position posSAT( 22300542.564, 9466839.117, 10798193.375 );

cout << posBCN << endl << posHANOI << endl << posHANOI.elevation( posSAT ) << " : " << posHANOI.azimuth( posSAT ) << endl;

GPSTk basics: RINEX

RinexObsStream rinexData("bahr1620.04o");

RinexObsHeader rinexHeader;rinexData >> rinexHeader;

cout << rinexHeader.markerName << " : " << rinexHeader.antType << " : " << rinexHeader.firstObs << endl;

RinexObsData obsData;

while( rinexData >> obsData ){ obsData.dump( cout );}

GPSTk basics: EphemerisSP3EphemerisStore SP3EphList;SP3EphList.loadFile( "igs13354.sp3" );

SatID sat( 10, SatID::systemGPS );

DayTime epoch( SP3EphList.getInitialTime() );DayTime epochEnd( SP3EphList.getFinalTime() );

while( epoch < epochEnd ) { epoch += 60.0; Xvt svPos;

try { svPos = SP3EphList.getXvt( sat, epoch ); cout << epoch.DOYsecond() << " " << svPos.x[0] << " " << svPos.x[1] << " " << svPos.x[2] << endl; } catch (InvalidRequest& e) { continue; }}

GPSTk basics: Ephemeris

GPSTk basics: Solid tidesPosition posHANOI;posHANOI.setGeodetic( 21.033, 105.85, 308.0 );

SolidTides solid;

DayTime epoch(2009,10,28,0,0,0.0);DayTime epochEnd(2009,10,29,0,0,0.0);

while( epoch < epochEnd ) { Triple hanoiTide; hanoiTide = solid.getSolidTide(epoch,posHANOI);

cout << epoch.DOYsecond() << " " << hanoiTide[0] << " " << hanoiTide[1] << " " << hanoiTide[2] << endl;

epoch += 300.0;}

GPSTk basics: Solid tides

GNSS Data Structures

• GNSS Data Structures (GDS) were designed to help in GNSS data processing.

• They address GNSS data management problems.

• You can program without them, but they will make your work a lot easier.

• Almost all GNSS data may be identified (or indexed) by:

– Source: Receiver logging data.– Satellite: GPS, GALILEO, GLONASS...– Epoch: Time the data belongs to.– Type: C1, P1, L2, etc, and other types.

Key point:Key point: Save Meta-Information: Save Meta-Information:

Save Save 4 indexes4 indexes to identify to identify

each piece of dataeach piece of data

• Source: Implemented as class SourceID.

• Satellite: Implemented as class SatID.

– Several systems: GPS, Galileo, Glonass...

• Epoch: Implemented as class DayTime.

• Type: Implemented as class TypeID.

– Includes basic observations: C1, P1, etc.

– Other “types”: Relativity, troposphere, etc.

– You can create new types as needed!!!.

GDS example

• GDS provide several methods to ease handling. For instance:

Keep only C1 observable in GDS:

gpsData.keepOnlyTypeID(TypeID::C1)

GDS example

• GDS provide several methods to ease handling. For instance:

Keep only C1 observable in GDS:

gpsData.keepOnlyTypeID(TypeID::C1)

GDS Method Type to keep

Implementation

• GDS provide several methods to ease

handling. For instance:

Remove a given satellite from GDS:

gpsData.removeSatID(sat21)

GDS Method Satellite to remove

Implementation

sat21 = SatID( 21, SatID::systemGPS )

Implementation

Access GDS in a matrix-like fashion:

gpsData.body(TypeID::LI)(sat14)

Implementation

Access GDS in a matrix-like fashion:

gpsData.body(TypeID::LI)(sat14)

GDS “Branches” SatelliteType

Processing paradigm

• Operator “>>” is overloaded:

We redefine it in C++ to make data“flow” from one processing step to the next

Processing paradigmExample:

Declare object “rinexFile” to take data out ofRinex observation file:

RinexObsStream rinexFile("ebre0300.02o");

Example:

Declare object “rinexFile” to take data out ofRinex observation file:

RinexObsStream rinexFile("ebre0300.02o");

Processing paradigm

Class to handleRinex Obs files

Object name Rinex file name

Declare object “gpsData”. This is a GDS:

gnssRinex gpsData;

Declare object “gpsData”. This is a GDS:

gnssRinex gpsData;

GDS Class Object name

Combine everything and take one epoch ofdata out of Rinex file into GDS:

RinexObsStream rinexFile("ebre0300.02o");gnssRinex gpsData;

rinexFile >> gpsData;

rinexFile >> gpsData; Processing line

rinexFile >> gpsData;

“Flow” operator

Processing line

Processing paradigm

• Main Idea: GNSS data processing becomes like an “assembly line”.– The GDS “flows” from one

“workstation” to the next, like a car in factory.

– And, like the car, the GDS changes along the processing line.

• For the developer, the GDS is like a “white box” holding all the information needed, and properly indexed.

Processing code-based GPS data with the GPSTk and GDS

A simple example:

C1 code observable +least-mean squares solver

Plain standard GPS processing

C1 + LMS

• Get data out of Rinex file, one epoch at a time– Get data into GDS (gpsData)– Do it until Rinex file ends.

C1 + LMS

while ( rinexFile >> gpsData ){ }

• Get data out of Rinex file, one epoch at a time– Get data into GDS (gpsData)– Do it until Rinex file ends.

C1 + LMS

while ( rinexFile >> gpsData ){ }

• Get data into modeling object gpsModel– GpsModel was previously initialized:

•Ephemeris, tropospheric model, Klobuchar ionospheric model, etc.

– Initialization phase is not shown here.

C1 + LMS

while ( rinexFile >> gpsData ){ gpsData >> gpsModel }

• Get data into modeling object gpsModel– GpsModel was previously initialized:

•Ephemeris, tropospheric model, Klobuchar ionospheric model, etc.

– Initialization phase is not shown here.

C1 + LMS

• Resulting GDS with original data and modeled data gets into solver object lmsSolver– lmsSolver is from SolverLMS class.– lmsSolver was previously initialized. – Initialization phase is not shown here.

C1 + LMS

while ( rinexFile >> gpsData ){ gpsData >> gpsModel >> lmsSolver; }

• Resulting GDS with original data and modeled data gets into solver object lmsSolver– lmsSolver is from SolverLMS class.– lmsSolver was previously initialized. – Initialization phase is not shown here.

C1 + LMS

while ( rinexFile >> gpsData ){ gpsData >> gpsModel >> lmsSolver; }

• Resulting GDS with original data and modeled data gets into solver object lmsSolver– lmsSolver is from SolverLMS class.– lmsSolver was previously initialized.– Initialization phase is not shown here.

Processingline

C1 + LMS

while ( rinexFile >> gpsData ){ gpsData >> gpsModel >> lmsSolver;

• All GNSS data processing is done!!!– Print results to the screen.– C++ cout object is used to print to screen.– Solution is inside lmsSolver: Ask for type.

C1 + LMS

while ( rinexFile >> gpsData ){ gpsData >> gpsModel >> lmsSolver;

cout << lmsSolver.getSolution(TypeID::dx) << '' ''; cout << lmsSolver.getSolution(TypeID::dy) << '' ''; cout << lmsSolver.getSolution(TypeID::dz) << endl;}

• All GNSS data processing is done!!!– Print results to the screen.– C++ cout object is used to print to screen.– Solution is inside lmsSolver: Ask for type.

C1 code observable +least-mean squares solver +East-North-Up (ENU) system

C1 + LMS + ENU

• Add a change-base object to processing line:– XYZ2NEU class change from ECEF to ENU.– Create object baseChange from XYZ2NEU

class.

C1 + LMS + ENU

while ( rinexFile >> gpsData ){ gpsData >> gpsModel >> baseChange }

• Add a change-base object to processing line:– XYZ2NEU class change from ECEF to NEU.– Create object baseChange from XYZ2NEU

class.

C1 + LMS + ENU

while ( rinexFile >> gpsData ){ gpsData >> gpsModel >> baseChange }

• Resulting GDS gets into object enuSolver– enuSolver belongs to SolverLMS class.– enuSolver was previously initialized to

solve using dLat, dLon,dUp and cdt.– Initialization phase is not shown here.

C1 + LMS + ENU

while ( rinexFile >> gpsData ){ gpsData >> gpsModel >> baseChange >> enuSolver; }

• Resulting GDS gets into object enuSolver– enuSolver belongs to SolverLMS class.– enuSolver was previously initialized to

solve using dLat, dLon,dUp and cdt.– Initialization phase is not shown here.

C1 + LMS + ENU

while ( rinexFile >> gpsData ){ gpsData >> gpsModel >> baseChange >> enuSolver;

• All GNSS data processing is done!!!– Print results to the screen.– Solution is inside enuSolver: Ask for type.

C1 + LMS + ENU

while ( rinexFile >> gpsData ){ gpsData >> gpsModel >> baseChange >> enuSolver;

cout << enuSolver.getSolution(TypeID::dLat) << '' ''; cout << enuSolver.getSolution(TypeID::dLon) << '' ''; cout << enuSolver.getSolution(TypeID::dH) << endl;}

• All GNSS data processing is done!!!– Print results to the screen.– Solution is inside enuSolver: Ask for type.

PC (ionosphere free) code + smoothing with LC + weighted least-mean squares solver + East-North-Up (ENU) system

Smoothed-PC + WMS + ENU

while ( rinexFile >> gpsData ){ gpsData >> }

• First, get combinations we will need:– PC and LC (ionosphere-free).– LI (ionospheric combination).– MW (Melbourne-Wübbena)

• Look for corresponding classes in GPSTk API

while ( rinexFile >> gpsData ){ gpsData >> getPC >> getLC >> getLI >> getMW}

• First, get combinations we will need:– PC and LC (ionosphere-free).– LI (ionospheric combination).– MW (Melbourne-Wübbena)

• Look for corresponding classes in GPSTk API

while ( rinexFile >> gpsData ){ gpsData >> getPC >> getLC >> getLI >> getMW

• Then, detect and mark cycle-slips:– Two different and complementary methods

are used.• Everything is in a single processing line.

>> markCSLI >> markCSMW

• Then, detect and mark cycle-slips:– Two different and complementary methods

are used.• Everything is in a single processing line.

>> markCSLI >> markCSMW

• Smooth code with phase

>> markCSLI >> markCSMW >> smoothPC

>> markCSLI >> markCSMW >> smoothPC>> gpsModel

• Apply model

>> markCSLI >> markCSMW >> smoothPC>> gpsModel >> getWeights

• Apply model

• Compute weights

>> markCSLI >> markCSMW >> smoothPC>> gpsModel >> getWeights >> baseChange>> wmsSolver;

• Apply model

• Compute weights

• Change base and solve with a WMS solver

>> markCSLI >> markCSMW >> smoothPC>> gpsModel >> getWeights >> baseChange>> wmsSolver;

• All GNSS data processing is done!!!– Print results to the screen.– Solution is inside wmsSolver: Ask for type.

• Everything is in a single processing line.

Differential GPS +C1 code observable +

weighted least-mean squares + East-North-Up (ENU) system

C1 + DGPS + WMS + ENU

while ( rinexFile >> gpsData ){ gpsDataRef >> synchro >> refModel;

• First, process reference station data:– synchro is a Synchronize class object.– It is in charge of synchronizing reference

and rover data streams.– Initialization is not shown here. Consult API.

delta.setRefData( gpsDataRef.body );

• Then, indicate the reference data:– delta is a DeltaOp class object.– It computes single diffences using common

satellites.– You must tell it which is the reference data.

gpsData >> gpsModel >> getWeights >> delta >> baseChange >> wmsSolver;

• Finally, process rover receiver data:– You must insert delta object into processing

line to compute single differences.

gpsData >> gpsModel >> getWeights >> delta >> baseChange >> wmsSolver;

• All GNSS data processing is done!!!– Print results to the screen.– Solution is inside wmsSolver: Ask for type.

Code-based results (as expected)

Processing phase-based GPS data with the GPSTk and GDS

Example ofPrecise Point Positioning (PPP)

PPP (static, forward mode)correctObservables.setExtraBiases(tides);

Set “correcting” object with tide information.It also includes RX antenna phase center and eccentricity

gpsData >> basicGPSModel

Compute basic components of model

gpsData >> basicGPSModel >> removeEclipsedSatellites

Remove satellites in eclipse

gpsData >> basicGPSModel >> removeEclipsedSatellites >> gravitationalDelay

Compute gravitational delay

gpsData >> basicGPSModel >> removeEclipsedSatellites >> gravitationalDelay >> computeSatellitePhaseCenter

Compute the effect of satellite phase centers

gpsData >> basicGPSModel >> removeEclipsedSatellites >> gravitationalDelay >> computeSatellitePhaseCenter >> correctObservables

Correct observables from tides, RX phase center, etc

gpsData >> basicGPSModel >> removeEclipsedSatellites >> gravitationalDelay >> computeSatellitePhaseCenter >> correctObservables >> windUp

Compute wind-up effect

gpsData >> basicGPSModel >> removeEclipsedSatellites >> gravitationalDelay >> computeSatellitePhaseCenter >> correctObservables >> windUp >> computeTropo

Compute tropospheric effect (Niell model)

gpsData >> basicGPSModel >> removeEclipsedSatellites >> gravitationalDelay >> computeSatellitePhaseCenter >> correctObservables >> windUp >> computeTropo >> computeLinearCombinations

Compute common linear combinations: PC, LC, LI, ...

gpsData >> basicGPSModel >> removeEclipsedSatellites >> gravitationalDelay >> computeSatellitePhaseCenter >> correctObservables >> windUp >> computeTropo >> computeLinearCombinations >> markCSLI >> markCSMW

Marck cycle slips: Two complementary algorithms

gpsData >> basicGPSModel >> removeEclipsedSatellites >> gravitationalDelay >> computeSatellitePhaseCenter >> correctObservables >> windUp >> computeTropo >> computeLinearCombinations >> markCSLI >> markCSMW >> computePrefitResiduals

Compute prefit residuals

gpsData >> basicGPSModel >> removeEclipsedSatellites >> gravitationalDelay >> computeSatellitePhaseCenter >> correctObservables >> windUp >> computeTropo >> computeLinearCombinations >> markCSLI >> markCSMW >> computePrefitResiduals >> decimateData

Decimate data if epoch is not a multiple of 900 seconds

gpsData >> basicGPSModel >> removeEclipsedSatellites >> gravitationalDelay >> computeSatellitePhaseCenter >> correctObservables >> windUp >> computeTropo >> computeLinearCombinations >> markCSLI >> markCSMW >> computePrefitResiduals >> decimateData >> baseChange

Change from ECEF to ENU reference frame

gpsData >> basicGPSModel >> removeEclipsedSatellites >> gravitationalDelay >> computeSatellitePhaseCenter >> correctObservables >> windUp >> computeTropo >> computeLinearCombinations >> markCSLI >> markCSMW >> computePrefitResiduals >> decimateData >> baseChange >> pppSolver;

Solve equations with a properly adjusted Kalman filter

gpsData >> basicGPSModel >> removeEclipsedSatellites >> gravitationalDelay >> computeSatellitePhaseCenter >> correctObservables >> windUp >> computeTropo >> computeLinearCombinations >> markCSLI >> markCSMW >> computePrefitResiduals >> decimateData >> baseChange >> pppSolver;

All this processing is repeated for each epoch

PPP: MADR 2008/05/27, static, forward

The achieved accuracy is better than 2 centimeters!!!

PPP (kinematic, forward mode)

WhiteNoiseModel newCoordinatesModel(100.0);

pppSolver.setCoordinatesModel(&newCoordinatesModel);

correctObservables.setExtraBiases(tides);

gpsData >> basicGPSModel >> removeEclipsedSatellites ... ... >> pppSolver;

PPP:MADR 2008/05/27, kinematic, forward

Comparison with other GPS software

Conclusions

• GPSTk is already a solid base to work upon, saving tedious work to the programmer and researcher.

• The GPSTk provides solid advantages both for companies and educational/research institutions.

• GPSTk portability, design and openness makes it very flexible for multiple uses.

• GDS provide a powerful, flexible and easy to use processing framework.

Conclusions

• GPSTk results are comparable with the best GPS data processing software available.

• There are several areas being improved:

– RINEX version 3 handling.

– Robust outlier detection classes.

– More sophisticated tide models.

– Other GNSS processing strategies (RTK).

Thanks for your attention and time!

gage using gpstk gnss learning research seagal hanoi 2009

Documents

vermont thread gage - special · pdf filevermont thread...

clare gage

starring steven seagal - saradan media · starring steven...

phineous gage

linear gage

vivo mare-seagal-subplot 5

gage kihlstrom

gage bibliography

g53a-0123 gpstk: an open source toolkit for working...

gage epistemology

gage studies

strain gage thermal output and gage factor variation · pdf...

index gage block gage blocks - mitutoyo america corporation

gage digitizer product overvie · 2016. 10. 24. · title:...

strain gage calibration factors for constant room...

cultuurpodiumboerderij.nlcultuurpodiumboerderij.nl/assets/cultuurpodium-boerderij-jaarverslag-2014.pdfrichie...

phineas gage

170707 스트레인게이지본딩 브로슈어 · strain...

gpsdp gage

propeller dimensional inspection report11. gage radial...