SolverComparer.cpp File Reference

Incremental and batch solving, timing, and accuracy comparisons. More...

Typedefs
typedef Pose2	Pose
typedef NoiseModelFactorN< Pose >	NM1
typedef NoiseModelFactorN< Pose, Pose >	NM2
typedef BearingRangeFactor< Pose, Point2 >	BR

Functions
double	chi2_red (const gtsam::NonlinearFactorGraph &graph, const gtsam::Values &config)
void	runIncremental ()
void	runBatch ()
void	runCompare ()
void	runPerturb ()
void	runStats ()
int	main (int argc, char *argv[])

Variables
string	outputFile
string	inputFile
string	datasetName
int	firstStep
int	lastStep
int	nThreads
int	relinSkip
bool	incremental
bool	dogleg
bool	batch
bool	compare
bool	perturb
bool	stats
double	perturbationNoise
string	compareFile1
string	compareFile2
Values	initial
NonlinearFactorGraph	datasetMeasurements

Detailed Description

Incremental and batch solving, timing, and accuracy comparisons.

Author

Richard Roberts

Date

August, 2013

Here is an example. Below, to run in batch mode, we first generate an initialization in incremental mode.

Solve in incremental and write to file w_inc: ./SolverComparer –incremental -d w10000 -o w_inc

You can then perturb that initialization to get batch something to optimize. Read in w_inc, perturb it with noise of stddev 0.6, and write to w_pert: ./SolverComparer –perturb 0.6 -i w_inc -o w_pert

Then optimize with batch, read in w_pert, solve in batch, and write to w_batch: ./SolverComparer –batch -d w10000 -i w_pert -o w_batch

And finally compare solutions in w_inc and w_batch to check that batch converged to the global minimum ./SolverComparer –compare w_inc w_batch