Generic interface for a sparse solver operating on a graph. More...

#include <solver.h>

Inheritance diagram for g2o::Solver:

Public Member Functions
size_t	additionalVectorSpace () const
double *	b ()
	return b, the right hand side of the system
const double *	b () const
virtual bool	buildStructure (bool zeroBlocks=false)=0
virtual bool	buildSystem ()=0
virtual bool	computeMarginals ()=0
virtual bool	computeMarginals (SparseBlockMatrix< MatrixXd > &spinv, const std::vector< std::pair< int, int > > &blockIndices)=0
virtual bool	init (bool online=false)=0
bool	levenberg () const
	the system is Levenberg-Marquardt
SparseOptimizer *	optimizer () const
	the optimizer (graph) on which the solver works
virtual bool	schur ()=0
	should the solver perform the schur complement or not
void	setAdditionalVectorSpace (size_t s)
virtual bool	setLambda (double lambda)=0
void	setLevenberg (bool levenberg)
void	setOptimizer (SparseOptimizer *optimizer)
virtual void	setSchur (bool s)=0
virtual bool	solve ()=0
	Solver (SparseOptimizer *optimizer)
virtual bool	supportsSchur ()
virtual bool	updateStructure (const std::vector< HyperGraph::Vertex * > &vset, const HyperGraph::EdgeSet &edges)=0
size_t	vectorSize () const
	return the size of the solution vector (x) and b
double *	x ()
	return x, the solution vector
const double *	x () const
virtual	~Solver ()
Protected Member Functions
void	resizeVector (size_t sx)
Protected Attributes
size_t	_additionalVectorSpace
double *	_b
bool	_isLevenberg
	the system we gonna solve is a Levenberg-Marquardt system
size_t	_maxXSize
SparseOptimizer *	_optimizer
double *	_x
size_t	_xSize

Detailed Description

Generic interface for a sparse solver operating on a graph.

Definition at line 33 of file solver.h.

Constructor & Destructor Documentation

g2o::Solver::Solver ( SparseOptimizer * optimizer ) [explicit]

Definition at line 24 of file solver.cpp.

g2o::Solver::~Solver ( ) [virtual]

Definition at line 30 of file solver.cpp.

Member Function Documentation

size_t g2o::Solver::additionalVectorSpace ( ) const [inline]

Definition at line 111 of file solver.h.

double* g2o::Solver::b ( ) [inline]

return b, the right hand side of the system

Definition at line 87 of file solver.h.

const double* g2o::Solver::b ( ) const [inline]

Definition at line 88 of file solver.h.

virtual bool g2o::Solver::buildStructure ( bool zeroBlocks = false ) [pure virtual]

build the structure of the system

Implemented in g2o::BlockSolver< Traits >.

virtual bool g2o::Solver::buildSystem ( ) [pure virtual]

build the current system

Implemented in g2o::BlockSolver< Traits >.

virtual bool g2o::Solver::computeMarginals ( ) [pure virtual]

computes the block diagonal elements of the inverted hessian and stores them in the nodes of the graph.

Implemented in g2o::BlockSolver< Traits >.

virtual bool g2o::Solver::computeMarginals	(	SparseBlockMatrix< MatrixXd > &	spinv,
		const std::vector< std::pair< int, int > > &	blockIndices
	)		`[pure virtual]`

computes the block diagonal elements of the pattern specified in the input and stores them in the nodes of the graph.

Implemented in g2o::BlockSolver< Traits >.

virtual bool g2o::Solver::init ( bool online = false ) [pure virtual]

initialize the solver, called once before the first iteration

Implemented in g2o::BlockSolver< Traits >.

bool g2o::Solver::levenberg ( ) const [inline]

the system is Levenberg-Marquardt

Definition at line 98 of file solver.h.

SparseOptimizer* g2o::Solver::optimizer ( ) const [inline]

the optimizer (graph) on which the solver works

Definition at line 94 of file solver.h.

void g2o::Solver::resizeVector ( size_t sx ) [protected]

Definition at line 36 of file solver.cpp.

virtual bool g2o::Solver::schur ( ) [pure virtual]

should the solver perform the schur complement or not

Implemented in g2o::BlockSolver< Traits >.

void g2o::Solver::setAdditionalVectorSpace ( size_t s )

Definition at line 72 of file solver.cpp.

virtual bool g2o::Solver::setLambda ( double lambda ) [pure virtual]

update the system while performing Levenberg, i.e., modifying the diagonal components of A by doing += lambda along the main diagonal of the Matrix. Note that this function may be called with a positive and a negative lambda. The latter is used to undo a former modification.

Implemented in g2o::BlockSolver< Traits >.

void g2o::Solver::setLevenberg ( bool levenberg )

Definition at line 67 of file solver.cpp.

void g2o::Solver::setOptimizer ( SparseOptimizer * optimizer )

Definition at line 62 of file solver.cpp.

virtual void g2o::Solver::setSchur ( bool s ) [pure virtual]

Implemented in g2o::BlockSolver< Traits >.

virtual bool g2o::Solver::solve ( ) [pure virtual]

solve Ax = b

Implemented in g2o::BlockSolver< Traits >.

virtual bool g2o::Solver::supportsSchur ( ) [inline, virtual]

does this solver support the Schur complement for solving a system consisting of poses and landmarks. Re-implemement in a derived solver, if your solver supports it.

Reimplemented in g2o::BlockSolver< Traits >.

Definition at line 105 of file solver.h.