Header-only framework for building Kalman-style filters, prediction, and sensor fusion. More...

Namespaces
	matrix_exponential_map
	Utilities for interacting with a "matrix exponential map vector" rotation parameterization/formalism, where rotation is represented as a 3D vector that is turned into a rotation matrix by applying Rodrigues' formula that resembles a matrix exponential.

	types
	Type aliases, including template type aliases.

Classes
class	AbsoluteOrientationBase
	The measurement here has been split into a base and derived type, so that the derived type only contains the little bit that depends on a particular state type. More...

class	AbsoluteOrientationMeasurement
	This is the subclass of AbsoluteOrientationBase: only explicit specializations, and on state types. More...

class	AbsoluteOrientationMeasurement< pose_externalized_rotation::State >
	AbsoluteOrientationMeasurement with a pose_externalized_rotation::State. More...

class	AbsolutePositionBase
	The measurement here has been split into a base and derived type, so that the derived type only contains the little bit that depends on a particular state type. More...

class	AbsolutePositionMeasurement
	This is the subclass of AbsolutePositionBase: only explicit specializations, and on state types. More...

class	AbsolutePositionMeasurement< pose_externalized_rotation::State >
	AbsolutePositionMeasurement with a pose_externalized_rotation::State. More...

class	AngularVelocityBase

class	AngularVelocityMeasurement
	This is the subclass of AngularVelocityBase: only explicit specializations, and on state types. More...

class	AngularVelocityMeasurement< orient_externalized_rotation::State >
	AngularVelocityMeasurement with a orient_externalized_rotation::State The code is in fact identical except for the state types, due to a coincidence of how the state vectors are arranged. More...

class	AngularVelocityMeasurement< pose_externalized_rotation::State >
	AngularVelocityMeasurement with a pose_externalized_rotation::State. More...

class	AugmentedProcessModel
	Process model type that consists entirely of references to two sub-process models, for operating on an AugmentedState<>. More...

class	AugmentedSigmaPointGenerator

class	AugmentedState
	State type that consists entirely of references to two independent sub-states. More...

class	ConstantProcess
	A simple process model for a "constant" process: all prediction does at most is bump up the uncertainty. More...

struct	CorrectionInProgress

class	FlexibleKalmanFilter
	The main class implementing the common components of the Kalman family of filters. More...

struct	HasDimension
	Convenience base class for things (like states and measurements) that have a dimension. More...

class	IMUAngVelMeasurement

class	IMUOrientationMeasBase
	The measurement here has been split into a base and derived type, so that the derived type only contains the little bit that depends on a particular state type. More...

class	IMUOrientationMeasurement
	This is the subclass of AbsoluteOrientationBase: only explicit specializations, and on state types. More...

class	IMUOrientationMeasurement< pose_externalized_rotation::State, PolicyT >
	AbsoluteOrientationMeasurement with a pose_externalized_rotation::State. More...

class	OrientationConstantVelocityProcessModel
	A model for a 3DOF pose (with angular velocity) More...

class	PoseConstantVelocityProcessModel
	A constant-velocity model for a 6DOF pose (with velocities) More...

class	PoseDampedConstantVelocityProcessModel
	A basically-constant-velocity model, with the addition of some damping of the velocities inspired by TAG. More...

class	PoseSeparatelyDampedConstantVelocityProcessModel
	A basically-constant-velocity model, with the addition of some damping of the velocities inspired by TAG. More...

class	PureVectorState
	A very simple (3D by default) vector state with no velocity, ideal for use as a position, with ConstantProcess for beacon autocalibration. More...

class	ReconstructedDistributionFromSigmaPoints

class	SigmaPointCorrectionApplication

struct	SigmaPointParameterDerivedQuantities

struct	SigmaPointParameters
	For further details on the scaling factors, refer to: Julier, S. More...

Typedefs
template<typename ModelA , typename ModelB >
using	DeducedAugmentedProcessModel = AugmentedProcessModel< typename std::remove_const< ModelA >::type, typename std::remove_const< ModelB >::type >
	Template alias to make removing const from the deduced types less verbose/painful. More...

template<typename StateA , typename StateB >
using	DeducedAugmentedState = AugmentedState< typename std::remove_const< StateA >::type, typename std::remove_const< StateB >::type >
	Template alias to make removing const from the deduced types less verbose/painful. More...

template<std::size_t Dim>
using	SigmaPointGenerator = AugmentedSigmaPointGenerator< Dim, Dim >

Functions
template<typename ModelA , typename ModelB >
DeducedAugmentedProcessModel< ModelA, ModelB >	makeAugmentedProcessModel (ModelA &a, ModelB &b)
	Factory function, akin to `std::tie()`, to make an augmented process model. More...

template<typename StateA , typename StateB >
DeducedAugmentedState< StateA, StateB >	makeAugmentedState (StateA &a, StateB &b)
	Factory function, akin to `std::tie()`, to make an augmented state.

template<typename StateType , typename ProcessModelType >
types::DimSquareMatrix< StateType >	predictErrorCovariance (StateType const &state, ProcessModelType &processModel, double dt)
	Computes P-. More...

template<typename StateType , typename ProcessModelType , typename MeasurementType >
CorrectionInProgress< StateType, MeasurementType >	beginCorrection (StateType &state, ProcessModelType &processModel, MeasurementType &meas)

template<typename StateType , typename ProcessModelType >
void	predict (StateType &state, ProcessModelType &processModel, double dt)

template<typename StateType , typename ProcessModelType , typename MeasurementType >
bool	correct (StateType &state, ProcessModelType &processModel, MeasurementType &meas, bool cancelIfNotFinite=true)

template<typename Derived >
Eigen::Matrix3d	makeSkewSymmetrixCrossProductMatrix (Eigen::MatrixBase< Derived > const &v)
	Produces the "hat matrix" that produces the same result as performing a cross-product with v. More...

template<typename State , typename Measurement >
SigmaPointCorrectionApplication< State, Measurement >	beginUnscentedCorrection (State &s, Measurement &m, SigmaPointParameters const &params=SigmaPointParameters())

double	computeNu (std::size_t L, double alpha)

Detailed Description

Header-only framework for building Kalman-style filters, prediction, and sensor fusion.

Typedef Documentation

§ DeducedAugmentedProcessModel

template<typename ModelA , typename ModelB >

using osvr::kalman::DeducedAugmentedProcessModel = typedef AugmentedProcessModel<typename std::remove_const<ModelA>::type, typename std::remove_const<ModelB>::type>

Template alias to make removing const from the deduced types less verbose/painful.

§ DeducedAugmentedState

template<typename StateA , typename StateB >

using osvr::kalman::DeducedAugmentedState = typedef AugmentedState<typename std::remove_const<StateA>::type, typename std::remove_const<StateB>::type>

Template alias to make removing const from the deduced types less verbose/painful.

Function Documentation

§ beginCorrection()

template<typename StateType , typename ProcessModelType , typename MeasurementType >

CorrectionInProgress<StateType, MeasurementType> osvr::kalman::beginCorrection	(	StateType &	state,
		ProcessModelType &	processModel,
		MeasurementType &	meas
	)

inline

Dimension of measurement

Dimension of state

Measurement Jacobian

Measurement covariance

State error covariance

The kalman gain stuff to not invert (called P12 in TAG)

the stuff to invert for the kalman gain also sometimes called S or the "Innovation Covariance"

More computation is done in initializers/constructor

§ correct()

template<typename StateType , typename ProcessModelType , typename MeasurementType >

bool osvr::kalman::correct	(	StateType &	state,
		ProcessModelType &	processModel,
		MeasurementType &	meas,
		bool	cancelIfNotFinite = `true`
	)

inline

Parameters

cancelIfNotFinite If the state correction or new error covariance is detected to contain non-finite values, should we cancel the correction and not apply it?

Returns: true if correction completed

§ makeAugmentedProcessModel()

template<typename ModelA , typename ModelB >

DeducedAugmentedProcessModel<ModelA, ModelB> osvr::kalman::makeAugmentedProcessModel	(	ModelA &	a,
		ModelB &	b
	)

inline

Factory function, akin to std::tie(), to make an augmented process model.

§ makeSkewSymmetrixCrossProductMatrix()

template<typename Derived >

Eigen::Matrix3d osvr::kalman::makeSkewSymmetrixCrossProductMatrix ( Eigen::MatrixBase< Derived > const & v )

inline

Produces the "hat matrix" that produces the same result as performing a cross-product with v.

This is the same as the "capital omega" skew-symmetrix matrix used by a matrix-exponential-map rotation vector.

Parameters

v	a 3D vector

Returns: a matrix M such that for some 3D vector u, Mu = v x u.

§ predictErrorCovariance()

template<typename StateType , typename ProcessModelType >

types::DimSquareMatrix<StateType> osvr::kalman::predictErrorCovariance	(	StateType const &	state,
		ProcessModelType &	processModel,
		double	dt
	)

inline

Computes P-.

Usage is optional, most likely called from the process model updateState()` method.

Todo:: Determine if the fact that Q is (at least in one case) symmetrical implies anything else useful performance-wise here or later in the data flow.

Namespaces

Classes

Typedefs

Functions

Detailed Description

Typedef Documentation

§ DeducedAugmentedProcessModel

§ DeducedAugmentedState

Function Documentation

§ beginCorrection()

§ correct()

§ makeAugmentedProcessModel()

§ makeSkewSymmetrixCrossProductMatrix()

§ predictErrorCovariance()