opentl::models::LinearARModels Class Reference

Implemented class Continuous White Noise Acceleration (CWNA) motion model. More...

Inherits opentl::models::Motion.

Inherited by opentl::models::Brownian, opentl::models::ConstantVelocity, and opentl::models::Oscillatory.

List of all members.

Public Member Functions
virtual math::SquareMatrix &	getNoiseCovariance ()
	Get process noise covariance matrix.
virtual int	getNoiseDim ()
	Returns dimension of process noise square matrix.
virtual void	getUniformPrior (opentl::core::State &srcDstState)
	Function to get the first state prior for the tracking. The filter calls this function at initialization. Requires to call setNoiseCovariance() first to get range values for uniform distribution.
	LinearARModels (MotionType motionType, boost::shared_ptr< opentl::core::State > stateTemplate, math::Matrix *matB=NULL, double fixedDeltaT=0.0)
	Constructor.
virtual void	predictMotion (opentl::core::State &srcDstState, bool addNoise=false, math::SquareMatrix **noiseCovMat=NULL, math::SquareMatrix **jacStateTrans=NULL, math::Vector *inputU=NULL)
	Predicts state(k+1) based on specific motion model.
virtual void	setAveragePose (const math::Vector &averagePose)
	Set average pose value (used by constrained models, with additive update only).
void	setFixedDeltaT (double fixedDeltaT=0.0f)
	Sets a fixed delta t.
virtual void	setNoiseCovariance (const math::SquareMatrix &noiseCovMat)
	Set process noise covariance, and update matrices accordingly.
virtual void	setUniformNoiseRange (const math::Vector &uniformNoiseRange)
	Set uniform noise range (for the initial prior).
virtual void	updateMatrices ()=0
	Update internal matrices (mMatA, mMatW, mMatQw) using the internal parameter mDeltaT.
virtual	~LinearARModels ()
	Destructor.
Protected Member Functions
void	convertMatricesToRealTime ()
	Convert AR matrices, from time-less (A,W) to real-time (At,Wt). Time-less state model: s(t) = [p(t),p(t-1)] Real-time state model: s(t) = [p(t),p_dot(t)] Conversion matrix: H = [I I*dt; I 0] New matrices: At = inv(H)*A*H; Wt = inv(H)*W;.
Protected Attributes
math::Matrix	matWTmp
math::Vector	mAveragePoseVec
	Average pose vector.
clock_t	mCurrentTime
double	mDeltaT
	Time difference (of old posterior state till now).
double	mFixedDeltaT
	Use a fixed time value for deltaT if this value is not -1.0.
math::SquareMatrix	mMatA
	Real-time state transition matrix (A).
math::SquareMatrix	mMatATimeless
	Time-less version of state transition matrix (A).
math::Matrix *	mMatB
	Control matrix (B) if input vector u is given.
math::SquareMatrix	mMatH
	Conversion matrix between time-less and real-time AR model (function of deltaT).
math::SquareMatrix	mMatNoiseCov
	Process noise covariance.
math::SquareMatrix	mMatQw
	Process noise covariance matrix Qw = (W*w)*(W*w)'.
math::Matrix	mMatW
	Real-time process noise matrix (W).
math::Matrix	mMatWTimeless
	Time-less version of process noise matrix (W).
math::Vector	mNoiseParamVecW
	Vectors containing std for Gaussian noise generation (obtained from mMatNoiseCov).
math::Vector	mNoiseVecW
	Process noise vector (w).
int	mPoseIncDim
	Pose degrees of freedom (increment dimension).
int	mStateIncDim
	Full state degrees of freedom (increment dimension).
math::Vector	mStateTmp
math::Vector	mUniformNoiseRange
	Range of uniform noise (-u,u) for initializing the *absolute* pose vector.
math::Vector	mVecTmp
	hold temporary state information

---

Detailed Description

Implemented class Continuous White Noise Acceleration (CWNA) motion model.

Author:

Erwin Roth, erwin.roth@weihenstephan.org; Giorgio Panin, panin@in.tum.de

---

Constructor & Destructor Documentation

opentl::models::LinearARModels::LinearARModels	(	MotionType	motionType,
		boost::shared_ptr< opentl::core::State >	stateTemplate,
		math::Matrix *	matB = NULL,
		double	fixedDeltaT = 0.0
	)

Constructor.

Parameters:

	motionType	Motion type used for Linear ARModels
	stateTemplate	The state representation this model is using.
	matB	Optional: input matrix B
	fixedDeltaT	Optional: (default=deltaT is dynamic).

virtual opentl::models::LinearARModels::~LinearARModels

(

)

[virtual]

Destructor.

---

Member Function Documentation

void opentl::models::LinearARModels::convertMatricesToRealTime

(

)

[protected]

Convert AR matrices, from time-less (A,W) to real-time (At,Wt). Time-less state model: s(t) = [p(t),p(t-1)] Real-time state model: s(t) = [p(t),p_dot(t)] Conversion matrix: H = [I I*dt; I 0] New matrices: At = inv(H)*A*H; Wt = inv(H)*W;.

virtual math::SquareMatrix& opentl::models::LinearARModels::getNoiseCovariance

(

)

[virtual]

Get process noise covariance matrix.

Implements opentl::models::Motion.

virtual int opentl::models::LinearARModels::getNoiseDim

(

)

[virtual]

Returns dimension of process noise square matrix.

Returns:

Dimension of process noise

Implements opentl::models::Motion.

virtual void opentl::models::LinearARModels::getUniformPrior

(

opentl::core::State &

srcDstState

)

[virtual]

Function to get the first state prior for the tracking. The filter calls this function at initialization. Requires to call setNoiseCovariance() first to get range values for uniform distribution.

Parameters:

srcDstState

State used as basis for initialization of motion

Implements opentl::models::Motion.

virtual void opentl::models::LinearARModels::predictMotion	(	opentl::core::State &	srcDstState,
		bool	addNoise = false,
		math::SquareMatrix **	noiseCovMat = NULL,
		math::SquareMatrix **	jacStateTrans = NULL,
		math::Vector *	inputU = NULL
	)			[virtual]

Predicts state(k+1) based on specific motion model.

Parameters:

	srcDstState	Variable holding input and output state
	addNoise	Optional: (default = false) if true, process noise will be added to state(k+1)
	noiseCovMat	Gives access to (dynamic) process noise covariance matrix
	jacStateTrans	Optional: returns Jacobian of state transition matrix
	inputU	Optional: input control vector (u)

Implements opentl::models::Motion.

virtual void opentl::models::LinearARModels::setAveragePose

(

const math::Vector &

averagePose

)

[inline, virtual]

Set average pose value (used by constrained models, with additive update only).

Parameters:

averagePose

Input average pose vector

Implements opentl::models::Motion.

void opentl::models::LinearARModels::setFixedDeltaT

(

double

fixedDeltaT = 0.0f

)

[virtual]

Sets a fixed delta t.

Parameters:

fixedDeltaT

value of delta t, if set to 0.0 then delta t is computed dynamically

Implements opentl::models::Motion.

virtual void opentl::models::LinearARModels::setNoiseCovariance

(

const math::SquareMatrix &

noiseCovMat

)

[virtual]

Set process noise covariance, and update matrices accordingly.

Implements opentl::models::Motion.

virtual void opentl::models::LinearARModels::setUniformNoiseRange

(

const math::Vector &

uniformNoiseRange

)

[inline, virtual]

Set uniform noise range (for the initial prior).

Implements opentl::models::Motion.

virtual void opentl::models::LinearARModels::updateMatrices

(

)

[pure virtual]

Update internal matrices (mMatA, mMatW, mMatQw) using the internal parameter mDeltaT.

Implemented in opentl::models::Brownian, opentl::models::ConstantVelocity, and opentl::models::Oscillatory.

---

Member Data Documentation

math::Matrix opentl::models::LinearARModels::matWTmp [protected]

math::Vector opentl::models::LinearARModels::mAveragePoseVec [protected]

Average pose vector.

clock_t opentl::models::LinearARModels::mCurrentTime [protected]

double opentl::models::LinearARModels::mDeltaT [protected]

Time difference (of old posterior state till now).

double opentl::models::LinearARModels::mFixedDeltaT [protected]

Use a fixed time value for deltaT if this value is not -1.0.

math::SquareMatrix opentl::models::LinearARModels::mMatA [protected]

Real-time state transition matrix (A).

math::SquareMatrix opentl::models::LinearARModels::mMatATimeless [protected]

Time-less version of state transition matrix (A).

math::Matrix* opentl::models::LinearARModels::mMatB [protected]

Control matrix (B) if input vector u is given.

math::SquareMatrix opentl::models::LinearARModels::mMatH [protected]

Conversion matrix between time-less and real-time AR model (function of deltaT).

math::SquareMatrix opentl::models::LinearARModels::mMatNoiseCov [protected]

Process noise covariance.

math::SquareMatrix opentl::models::LinearARModels::mMatQw [protected]

Process noise covariance matrix Qw = (W*w)*(W*w)'.

math::Matrix opentl::models::LinearARModels::mMatW [protected]

Real-time process noise matrix (W).

math::Matrix opentl::models::LinearARModels::mMatWTimeless [protected]

Time-less version of process noise matrix (W).

math::Vector opentl::models::LinearARModels::mNoiseParamVecW [protected]

Vectors containing std for Gaussian noise generation (obtained from mMatNoiseCov).

math::Vector opentl::models::LinearARModels::mNoiseVecW [protected]

Process noise vector (w).

int opentl::models::LinearARModels::mPoseIncDim [protected]

Pose degrees of freedom (increment dimension).

int opentl::models::LinearARModels::mStateIncDim [protected]

Full state degrees of freedom (increment dimension).

math::Vector opentl::models::LinearARModels::mStateTmp [protected]

math::Vector opentl::models::LinearARModels::mUniformNoiseRange [protected]

Range of uniform noise (-u,u) for initializing the *absolute* pose vector.

math::Vector opentl::models::LinearARModels::mVecTmp [protected]

hold temporary state information