The root class for all feature models. More...

Detailed Description

The root class for all feature models.

It adds to the generic BlockObserveModel specific things related to slam. Especially it takes into account all the frame composition induced by robot pose and sensor pose relative pose on robot.

Definition at line 99 of file featureModel.hpp.

#include <featureModel.hpp>

Inheritance diagram for jafar::slam::BaseFeatureObserveModel:

[legend]

List of all members.

Public Member Functions
	BaseFeatureObserveModel (FeatureModel &featureModel, std::size_t sizeObs_)
	Constructor for systems with equal init and update observe models.
	BaseFeatureObserveModel (FeatureModel &featureModel_, std::size_t sizeObs_, std::size_t sizeObsInit_)
	Constructor for systems with differing init and update observe models.
	BaseFeatureObserveModel (FeatureModel &featureModel_, std::size_t sizeObs_, std::size_t sizeObsInit_, std::size_t sizeInnovation_, std::size_t sizePrediction)
	Constructor for systems with differing init and update observe models.
std::size_t	sizeRobotPose () const
std::size_t	sizeState () const
virtual Observation::ObservationType	typeObs () const =0
jafar::geom::T3D const &	robotToSensor () const
void	setRobotToSensor (jafar::geom::T3DEuler const &robotToSensor_)
	Set the position of the sensor within the robot frame.
virtual jblas::sym_mat const &	getR () const
	Get observation noise covariances matrix.
std::size_t	sizeObsInit ()
	Get size of RInit.
virtual jblas::sym_mat const &	getRInit () const
	Get init observation noise covariances matrix.
virtual void	computeR (Observation const &obs)
	Compute observation noise covariance.
virtual void	computeRInit (Observation const &obs)
	Compute observation noise covariance for init purposes.
virtual void	computeSensorR (Observation const &obs)
	Compute observation noise from sensor.
virtual void	computeSensorRInit (Observation const &obs)
	Compute sensor observation noise for initialization purposes.
virtual jblas::vec const &	predictObservation (const jblas::vec_range &robotPose_, const jblas::vec_range &feature_)
	Predict observation from robot and feature state.
virtual void	predictObservationJac (const jblas::vec_range &robotPose_, const jblas::vec_range &feature_)
	Computes jacobian of predictObservation().
virtual void	initInternalState (BaseFeature &feature, jafar::geom::T3DEuler const &robotPose_, Observation const &obs_) const
	Init internal state.
virtual void	updateInternalState (BaseFeature &feature, jafar::geom::T3DEuler const &robotPose_, Observation const &obs_) const
	Update internal state.
Public Attributes
FeatureModel &	featureModel
jblas::sym_mat	RInit
	Observation covariances matrix for initialization purposes.
jafar::jmath::GaussianVector	zPred
	Mean and covariance of the expected measurement.
Protected Member Functions
void	computePredictObservationJac (const jblas::vec_range &robotPose_, const jblas::vec_range &feature_, jblas::mat &J1, jblas::mat &J2)
	Compute prediction observation jacobian.
virtual jblas::vec const &	predictObservationInSensorFrame (jblas::vec const &featureInSensorFrame_)=0
	Prediction of the observation, the feature is expressed in the sensor frame featureInSensorFrame_.
virtual void	predictObservationInSensorFrameJac (jblas::vec const &featureInSensorFrame_)=0
	Jacobian of predictObservationInSensorFrame().
Protected Attributes
jblas::mat	JobsSensor
	jacobian of predictObservationInSensorFrame() function
jafar::geom::T3D *	m_robotToSensor
	Position of the sensor within the robot frame.
jblas::mat	JobsRobotToSensor
	jacobian of predictObservation() with respect to robotToSensor
jblas::sym_mat	featureR
	Observe model noise covariance.
std::size_t	p_sizeObsInit
	Size of obsevations for initialization purposes.

Member Function Documentation

virtual void jafar::slam::BaseFeatureObserveModel::computeR ( Observation const & obs ) [virtual]

Compute observation noise covariance.

It takes into account robotToSensor uncertainty.

virtual void jafar::slam::BaseFeatureObserveModel::computeRInit ( Observation const & obs ) [virtual]

Compute observation noise covariance for init purposes.

It should consider uncertainty in robotToSensor() <- TODO: not implemented.

virtual void jafar::slam::BaseFeatureObserveModel::computeSensorR ( Observation const & obs ) [inline, virtual]

Compute observation noise from sensor.

A default implementation is provided, it does nothing (constant model).

Reimplemented in jafar::slam::StereoImagePluckerFeatureObserveModel, jafar::slam::ScaledHomogeneousImagePluckerFeatureObserveModel, jafar::slam::RhoThetaImagePluckerFeatureObserveModel, jafar::slam::ImageEuclideanPluckerFeatureObserveModel, jafar::slam::ImageAnchorPluckerFeatureObserveModel, jafar::slam::ImageSegInvDepthFeatureObserveModel, jafar::slam::ImageSegInvDepthInovFeatureObserveModel, and jafar::slam::PolarPointFeatureObserveModel.

Definition at line 162 of file featureModel.hpp.

Referenced by computeSensorRInit().

virtual jblas::sym_mat const& jafar::slam::BaseFeatureObserveModel::getRInit ( ) const [inline, virtual]

Get init observation noise covariances matrix.

The init observation is the observation used for initialization purposes. It normally coincides with the observation noise, but not necessarily> it depends on the landmark type and the chosen parametrizations of the measurement space at initialization (feature detection) and update (feature matching) times.

Definition at line 145 of file featureModel.hpp.

References RInit.

virtual void jafar::slam::BaseFeatureObserveModel::initInternalState	(	BaseFeature &	feature,
		jafar::geom::T3DEuler const &	robotPose_,
		Observation const &	obs_
	)		const `[inline, virtual]`

Init internal state.

It is a set of parameters, describing the feature, but not integrated to the SLAM stochastic map.

Reimplemented in jafar::slam::StereoImagePluckerFeatureObserveModel, jafar::slam::ImagePluckerFeatureObserveModel, jafar::slam::ImageEuclideanPluckerFeatureObserveModel, jafar::slam::ImageAnchorPluckerFeatureObserveModel, jafar::slam::ImageSegInvDepthFeatureObserveModel, and jafar::slam::ImageSegInvDepthInovFeatureObserveModel.

Definition at line 190 of file featureModel.hpp.

virtual jblas::vec const& jafar::slam::BaseFeatureObserveModel::predictObservation	(	const jblas::vec_range &	robotPose_,
		const jblas::vec_range &	feature_
	)		`[virtual]`

Predict observation from robot and feature state.

Some work is done considering robot and sensor frames, and observe model is delegated until call to predictObservationInSensorFrame().

Implements jafar::filter::BlockObserveModel.

virtual void jafar::slam::BaseFeatureObserveModel::predictObservationJac	(	const jblas::vec_range &	robotPose_,
		const jblas::vec_range &	feature_
	)		`[virtual]`

Computes jacobian of predictObservation().

Same consideration as for predictObservation() and call to predictObservationInSensorFrameJac().

Implements jafar::filter::BlockObserveModel.

Reimplemented in jafar::slam::ImageSegInvDepthInovFeatureObserveModel.

virtual void jafar::slam::BaseFeatureObserveModel::updateInternalState	(	BaseFeature &	feature,
		jafar::geom::T3DEuler const &	robotPose_,
		Observation const &	obs_
	)		const `[inline, virtual]`

Update internal state.

See also:: initInternalState()

Reimplemented in jafar::slam::StereoImagePluckerFeatureObserveModel, jafar::slam::ImagePluckerFeatureObserveModel, jafar::slam::ImageEuclideanPluckerFeatureObserveModel, jafar::slam::ImageAnchorPluckerFeatureObserveModel, jafar::slam::ImageSegInvDepthFeatureObserveModel, and jafar::slam::ImageSegInvDepthInovFeatureObserveModel.

Definition at line 197 of file featureModel.hpp.

Member Data Documentation

jblas::sym_mat jafar::slam::BaseFeatureObserveModel::featureR [protected]

Observe model noise covariance.

It includes observation noise plus robotToSensor transformation uncertainty.

Definition at line 225 of file featureModel.hpp.

Referenced by getR().

jblas::sym_mat jafar::slam::BaseFeatureObserveModel::RInit

Observation covariances matrix for initialization purposes.

The init observation is the observation used for initialization purposes. It normally coincides with the observation noise, but not necessarily> it depends on the landmark type and the chosen parametrizations of the measurement space at initialization (feature detection) and update (feature matching) times.

Definition at line 122 of file featureModel.hpp.

Referenced by computeSensorRInit(), and getRInit().

The documentation for this class was generated from the following file:

modules/slam/include/slam/featureModel.hpp

Detailed Description

Public Member Functions

Public Attributes

Protected Member Functions

Protected Attributes

Member Function Documentation

Member Data Documentation