GPS ToolKit Software Library: SolverPPP Class Reference

This class may be used either in a Vector- and Matrix-oriented way, or with GNSS data structure objects from "DataStructures" class (much more simple to use it this way).

      // INITIALIZATION PART

      // EBRE station nominal position
   Position nominalPos(4833520.192, 41537.1043, 4147461.560);
   RinexObsStream rin("ebre0300.02o");  // Data stream

      // Load all the SP3 ephemerides files
   SP3EphemerisStore SP3EphList;
   SP3EphList.loadFile("igs11512.sp3");
   SP3EphList.loadFile("igs11513.sp3");
   SP3EphList.loadFile("igs11514.sp3");

   NeillTropModel neillTM( nominalPos.getAltitude(),
                           nominalPos.getGeodeticLatitude(),
                           30 );

      // Objects to compute the tropospheric data
   BasicModel basicM(nominalPos, SP3EphList);
   ComputeTropModel computeTropo(neillTM);

      // More declarations here: ComputeMOPSWeights, SimpleFilter,
      // LICSDetector, MWCSDetector, SolidTides, OceanLoading, 
      // PoleTides, CorrectObservables, ComputeWindUp, ComputeLinear,
      // LinearCombinations, etc.

      // Declare a SolverPPP object
   SolverPPP pppSolver;

     // PROCESSING PART

   gnssRinex gRin;

   while(rin >> gRin)
   {
      try
      {
         gRin  >> basicM
               >> correctObs
               >> compWindup
               >> computeTropo
               >> linear1      // Compute combinations
               >> pcFilter
               >> markCSLI2
               >> markCSMW
               >> markArc
               >> linear2      // Compute prefit residuals
               >> phaseAlign
               >> pppSolver;
      }
      catch(...)
      {
         cerr << "Unknown exception at epoch: " << time << endl;
         continue;
      }

         // Print results
      cout << time.DOYsecond()      << "  "; // Output field #1
      cout << pppSolver.solution[1] << "  "; // dx: Output field #2
      cout << pppSolver.solution[2] << "  "; // dy: Output field #3
      cout << pppSolver.solution[3] << "  "; // dz: Output field #4
      cout << pppSolver.solution[0] << "  "; // wetTropo: Out field #5
      cout << endl;
   }

The "SolverPPP" object will extract all the data it needs from the GNSS data structure that is "gRin" and will try to solve the PPP system of equations using a Kalman filter. It will also insert back postfit residual data (both code and phase) into "gRin" if it successfully solves the equation system.

By default, it will build the geometry matrix from the values of coefficients wetMap, dx, dy, dz and cdt, IN THAT ORDER. Please note that the first field of the solution will be the estimation of the zenital wet tropospheric component (or at least, the part that wasn't modeled by the tropospheric model used).

You may configure the solver to work with a NEU system in the class constructor or using the "setNEU()" method.

In any case, the "SolverPPP" object will also automatically add and estimate the ionosphere-free phase ambiguities. The independent vector is composed of the code and phase prefit residuals.

This class expects some weights assigned to each satellite. That can be achieved with objects from classes such as "ComputeIURAWeights", "ComputeMOPSWeights", etc.

If these weights are not assigned, then the "SolverPPP" object will set a value of "1.0" to code measurements, and "weightFactor" to phase measurements. The default value of "weightFactor" is "10000.0". This implies that code sigma is 1 m, and phase sigma is 1 cm.

For instance, in orden to use a 'full kinematic' mode we assign a white noise model to all the coordinates:

      // Define a white noise model with 100 m of sigma
   WhiteNoiseModel wnM(100.0);

      // Configure the solver to use this model for all coordinates
   pppSolver.setCoordinatesModel(&wnM);

Be aware, however, that you MUST NOT use this method to set a state-aware stochastic model (like RandomWalkModel, for instance) to ALL coordinates, because the results will certainly be erroneous. Use this method ONLY with non-state-aware stochastic models like 'StochasticModel' (constant coordinates) or 'WhiteNoiseModel'.

In order to overcome the former limitation, this class provides methods to set different, specific stochastic models for each coordinate, like:

      // Define a white noise model with 2 m of sigma for horizontal
      // coordinates (in this case, the solver is previously set to use
      // dLat, dLon and dH).
   WhiteNoiseModel wnHorizontalModel(2.0);

      // Define a random walk model with 0.04 m*m/s of process spectral
      // density for vertical coordinates.
   RandomWalkModel rwVerticalModel(0.04);

      // Configure the solver to use these models
   pppSolver.setXCoordinatesModel(&wnHorizontalModel);
   pppSolver.setYCoordinatesModel(&wnHorizontalModel);
   pppSolver.setZCoordinatesModel(&rwVerticalModel);

Definition at line 210 of file SolverPPP.hpp.


Public Member Functions
	SolverPPP (bool useNEU=false)
	Common constructor.
virtual int	Compute (const Vector< double > &prefitResiduals, const Matrix< double > &designMatrix, const Matrix< double > &weightMatrix) throw (InvalidSolver)
	Compute the PPP Solution of the given equations set.
virtual int	Compute (const Vector< double > &prefitResiduals, const Matrix< double > &designMatrix, const Vector< double > &weightVector) throw (InvalidSolver)
	Compute the PPP Solution of the given equations set.
virtual gnssSatTypeValue &	Process (gnssSatTypeValue &gData) throw (ProcessingException)
	Returns a reference to a gnnsSatTypeValue object after solving the previously defined equation system.
virtual gnssRinex &	Process (gnssRinex &gData) throw (ProcessingException)
	Returns a reference to a gnnsRinex object after solving the previously defined equation system.
virtual SolverPPP &	Reset (const Vector< double > &newState, const Matrix< double > &newErrorCov)
	Resets the PPP internal Kalman filter.
virtual SolverPPP &	setNEU (bool useNEU)
	Sets if a NEU system will be used.
virtual double	getWeightFactor (void) const
	Get the weight factor multiplying the phase measurements sigmas.
virtual SolverPPP &	setWeightFactor (double factor)
	Set the weight factor multiplying the phase measurement sigma.
StochasticModel *	getXCoordinatesModel () const
	Get stochastic model pointer for dx (or dLat) coordinate.
SolverPPP &	setXCoordinatesModel (StochasticModel *pModel)
	Set coordinates stochastic model for dx (or dLat) coordinate.
StochasticModel *	getYCoordinatesModel () const
	Get stochastic model pointer for dy (or dLon) coordinate.
SolverPPP &	setYCoordinatesModel (StochasticModel *pModel)
	Set coordinates stochastic model for dy (or dLon) coordinate.
StochasticModel *	getZCoordinatesModel () const
	Get stochastic model pointer for dz (or dH) coordinate.
SolverPPP &	setZCoordinatesModel (StochasticModel *pModel)
	Set coordinates stochastic model for dz (or dH) coordinate.
virtual SolverPPP &	setCoordinatesModel (StochasticModel *pModel)
	Set a single coordinates stochastic model to ALL coordinates.
virtual StochasticModel *	getTroposphereModel (void) const
	Get wet troposphere stochastic model pointer.
virtual SolverPPP &	setTroposphereModel (StochasticModel *pModel)
	Set zenital wet troposphere stochastic model.
virtual StochasticModel *	getReceiverClockModel (void) const
	Get receiver clock stochastic model pointer.
virtual SolverPPP &	setReceiverClockModel (StochasticModel *pModel)
	Set receiver clock stochastic model.
virtual StochasticModel *	getPhaseBiasesModel (void) const
	Get phase biases stochastic model pointer.
virtual SolverPPP &	setPhaseBiasesModel (StochasticModel *pModel)
	Set phase biases stochastic model.
virtual Matrix< double >	getPhiMatrix (void) const
	Get the State Transition Matrix (phiMatrix).
virtual SolverPPP &	setPhiMatrix (const Matrix< double > &pMatrix)
	Set the State Transition Matrix (phiMatrix).
virtual Matrix< double >	getQMatrix (void) const
	Get the Noise covariance matrix (QMatrix).
virtual SolverPPP &	setQMatrix (const Matrix< double > &pMatrix)
	Set the Noise Covariance Matrix (QMatrix).
virtual SolverPPP &	setKinematic (bool kinematicMode=true, double sigmaX=100.0, double sigmaY=100.0, double sigmaZ=100.0)
	Set the positioning mode, kinematic or static.
virtual int	getIndex (void) const
	Returns an index identifying this object.
virtual std::string	getClassName (void) const
	Returns a string identifying this object.
virtual	~SolverPPP ()
	Destructor.

Constructor & Destructor Documentation

SolverPPP ( bool useNEU = false )

Common constructor.

Parameters:

useNEU

If true, will compute dLat, dLon, dH coordinates; if false (the default), will compute dx, dy, dz.

Definition at line 57 of file SolverPPP.cpp.

References SolverPPP::setNEU().

virtual ~SolverPPP ( ) [inline, virtual]

Destructor.

Definition at line 493 of file SolverPPP.hpp.

Member Function Documentation

int Compute	(	const Vector< double > &	prefitResiduals,
		const Matrix< double > &	designMatrix,
		const Vector< double > &	weightVector
	)	throw (InvalidSolver) `[virtual]`

Compute the PPP Solution of the given equations set.

Parameters:

	prefitResiduals	Vector of prefit residuals
	designMatrix	Design matrix for the equation system
	weightVector	Vector of weights assigned to each satellite.

Warning:: A typical Kalman filter works with the measurements noise covariance matrix, instead of the vector of weights. Beware of this detail, because this method uses the later.

Returns:: 0 if OK -1 if problems arose

Reimplemented from CodeKalmanSolver.

Definition at line 121 of file SolverPPP.cpp.

References GPSTK_THROW.

int Compute	(	const Vector< double > &	prefitResiduals,
		const Matrix< double > &	designMatrix,
		const Matrix< double > &	weightMatrix
	)	throw (InvalidSolver) `[virtual]`

Compute the PPP Solution of the given equations set.

Parameters:

	prefitResiduals	Vector of prefit residuals
	designMatrix	Design matrix for the equation system
	weightMatrix	Matrix of weights

Warning:: A typical Kalman filter works with the measurements noise covariance matrix, instead of the matrix of weights. Beware of this detail, because this method uses the later.

Returns:: 0 if OK -1 if problems arose

Reimplemented from CodeKalmanSolver.

Definition at line 172 of file SolverPPP.cpp.

References GPSTK_RETHROW, GPSTK_THROW, and gpstk::inverseChol().

std::string getClassName ( void ) const [virtual]

Returns a string identifying this object.

Reimplemented from CodeKalmanSolver.

Reimplemented in SolverPPPFB.

Definition at line 48 of file SolverPPP.cpp.

int getIndex ( void ) const [virtual]

Returns an index identifying this object.

Reimplemented from CodeKalmanSolver.

Reimplemented in SolverPPPFB.

Definition at line 43 of file SolverPPP.cpp.

virtual StochasticModel* getPhaseBiasesModel ( void ) const [inline, virtual]

Get phase biases stochastic model pointer.

Definition at line 418 of file SolverPPP.hpp.

virtual Matrix<double> getPhiMatrix ( void ) const [inline, virtual]

Get the State Transition Matrix (phiMatrix).

Reimplemented from CodeKalmanSolver.

Definition at line 440 of file SolverPPP.hpp.

virtual Matrix<double> getQMatrix ( void ) const [inline, virtual]

Get the Noise covariance matrix (QMatrix).

Reimplemented from CodeKalmanSolver.

Definition at line 458 of file SolverPPP.hpp.

virtual StochasticModel* getReceiverClockModel ( void ) const [inline, virtual]

Get receiver clock stochastic model pointer.

Reimplemented from CodeKalmanSolver.

Definition at line 399 of file SolverPPP.hpp.

virtual StochasticModel* getTroposphereModel ( void ) const [inline, virtual]

Get wet troposphere stochastic model pointer.

Definition at line 380 of file SolverPPP.hpp.

virtual double getWeightFactor ( void ) const [inline, virtual]

Get the weight factor multiplying the phase measurements sigmas.

This factor is the code_sigma/phase_sigma ratio.

Definition at line 307 of file SolverPPP.hpp.

References gpstk::sqrt().

StochasticModel* getXCoordinatesModel ( ) const [inline]

Get stochastic model pointer for dx (or dLat) coordinate.

Reimplemented from CodeKalmanSolver.

Definition at line 324 of file SolverPPP.hpp.

StochasticModel* getYCoordinatesModel ( ) const [inline]

Get stochastic model pointer for dy (or dLon) coordinate.

Reimplemented from CodeKalmanSolver.

Definition at line 338 of file SolverPPP.hpp.

StochasticModel* getZCoordinatesModel ( ) const [inline]

Get stochastic model pointer for dz (or dH) coordinate.

Reimplemented from CodeKalmanSolver.

Definition at line 352 of file SolverPPP.hpp.

virtual SolverPPP& Reset	(	const Vector< double > &	newState,
		const Matrix< double > &	newErrorCov
	)	`[inline, virtual]`

Resets the PPP internal Kalman filter.

Parameters:

	newState	System state vector
	newErrorCov	Error covariance matrix

Warning:: Take care of dimensions: In this case newState must be 6x1 and newErrorCov must be 6x6.

Definition at line 290 of file SolverPPP.hpp.

SolverPPP & setCoordinatesModel ( StochasticModel * pModel ) [virtual]

Set a single coordinates stochastic model to ALL coordinates.

Parameters:

pModel

Pointer to StochasticModel associated with coordinates.

Warning:: Do NOT use this method to set the SAME state-aware stochastic model (like RandomWalkModel, for instance) to ALL coordinates, because the results will certainly be erroneous. Use this method only with non-state-aware stochastic models like 'StochasticModel' (constant coordinates) or 'WhiteNoiseModel'.

Reimplemented from CodeKalmanSolver.

Definition at line 774 of file SolverPPP.cpp.

Referenced by SolverPPP::setKinematic().

virtual SolverPPP& setPhaseBiasesModel ( StochasticModel * pModel ) [inline, virtual]

Set phase biases stochastic model.

Parameters:

pModel

Pointer to StochasticModel associated with phase biases.

Warning:: Be aware that some stochastic models store their internal state (for instance, 'RandomWalkModel' and 'PhaseAmbiguityModel'). If that is your case, you MUST NOT use the SAME model in DIFFERENT solver objects.
This method should be used with caution, because model must be of PhaseAmbiguityModel class in order to make sense.

Definition at line 435 of file SolverPPP.hpp.

virtual SolverPPP& setPhiMatrix ( const Matrix< double > & pMatrix ) [inline, virtual]

Set the State Transition Matrix (phiMatrix).

Parameters:

pMatrix

State Transition matrix.

Warning:: Process() methods set phiMatrix and qMatrix according to the stochastic models already defined. Therefore, you must use the Compute() methods directly if you use this method.

Reimplemented from CodeKalmanSolver.

Definition at line 453 of file SolverPPP.hpp.

virtual SolverPPP& setQMatrix ( const Matrix< double > & pMatrix ) [inline, virtual]

Set the Noise Covariance Matrix (QMatrix).

Parameters:

pMatrix

Noise Covariance matrix.

Warning:: Process() methods set phiMatrix and qMatrix according to the stochastic models already defined. Therefore, you must use the Compute() methods directly if you use this method.

Reimplemented from CodeKalmanSolver.

Definition at line 471 of file SolverPPP.hpp.

virtual SolverPPP& setReceiverClockModel ( StochasticModel * pModel ) [inline, virtual]

Set receiver clock stochastic model.

Parameters:

pModel

Pointer to StochasticModel associated with receiver clock.

Warning:: Be aware that some stochastic models store their internal state (for instance, 'RandomWalkModel' and 'PhaseAmbiguityModel'). If that is your case, you MUST NOT use the SAME model in DIFFERENT solver objects.

Reimplemented from CodeKalmanSolver.

Definition at line 413 of file SolverPPP.hpp.

virtual SolverPPP& setTroposphereModel ( StochasticModel * pModel ) [inline, virtual]

Set zenital wet troposphere stochastic model.

Parameters:

pModel

Pointer to StochasticModel associated with zenital wet troposphere.

Warning:: Be aware that some stochastic models store their internal state (for instance, 'RandomWalkModel' and 'PhaseAmbiguityModel'). If that is your case, you MUST NOT use the SAME model in DIFFERENT solver objects.

Definition at line 394 of file SolverPPP.hpp.

virtual SolverPPP& setWeightFactor ( double factor ) [inline, virtual]

Set the weight factor multiplying the phase measurement sigma.

Parameters:

factor

Factor multiplying the phase measurement sigma

Warning:: This factor should be the code_sigma/phase_sigma ratio. For instance, if we assign a code sigma of 1 m and a phase sigma of 10 cm, the ratio is 100, and so should be "factor".

Definition at line 319 of file SolverPPP.hpp.

SolverPPP Class Reference
[Mathematical algorithms, GPS solution algorithms and Tropospheric]

Detailed Description

Public Member Functions

Constructor & Destructor Documentation

Member Function Documentation

SolverPPP Class Reference [Mathematical algorithms, GPS solution algorithms and Tropospheric]

Detailed Description

Public Member Functions

Constructor & Destructor Documentation

Member Function Documentation

SolverPPP Class Reference
[Mathematical algorithms, GPS solution algorithms and Tropospheric]