#include <kalman_filter.h>

Public Member Functions
	KalmanFilter (unsigned int ns, unsigned int nm, unsigned int nc, const vnl_matrix< double > &Ai, const vnl_matrix< double > &Hi, const vnl_matrix< double > &Bi, const vnl_matrix< double > &z_initial, const vnl_matrix< double > &x_initial, const vnl_matrix< double > &Pi)
	Constructor.
	KalmanFilter (unsigned int ns, unsigned int nm, const vnl_matrix< double > &Ai, const vnl_matrix< double > &Hi, const vnl_matrix< double > &z_initial, const vnl_matrix< double > &x_initial, const vnl_matrix< double > &Pi)
	Constructor.
void	set_initial_input (const vnl_matrix< double > &x_initial)
void	measurement_update (const vnl_matrix< double > &zk, const vnl_matrix< double > &Rk)
	measurement_update.
vnl_matrix< double >	predict (const vnl_matrix< double > &Qk)
	predict.
vnl_matrix< double >	predict (const vnl_matrix< double > &Qk, const vnl_matrix< double > &uk)
	predict.
vnl_matrix< double >	update_predict (const vnl_matrix< double > &zk, const vnl_matrix< double > &Rk, const vnl_matrix< double > &Qk)
	update_predict.
vnl_matrix< double >	estimate () const
vnl_matrix< double >	prediction () const
Private Attributes
unsigned int	num_signal_dimensions
unsigned int	num_measurement_dimensions
unsigned int	num_control_dimensions
vnl_matrix< double >	A
vnl_matrix< double >	H
vnl_matrix< double >	B
vnl_matrix< double >	x
vnl_matrix< double >	x_pred
vnl_matrix< double >	z
vnl_matrix< double >	P
vnl_matrix< double >	K
Friends
vcl_ostream &	operator<< (vcl_ostream &os, const KalmanFilter &kf)
	output operator.

Detailed Description

Definition at line 36 of file kalman_filter.h.

Constructor & Destructor Documentation

KalmanFilter::KalmanFilter	(	unsigned int	ns,
		unsigned int	nm,
		unsigned int	nc,
		const vnl_matrix< double > &	Ai,
		const vnl_matrix< double > &	Hi,
		const vnl_matrix< double > &	Bi,
		const vnl_matrix< double > &	z_initial,
		const vnl_matrix< double > &	x_initial,
		const vnl_matrix< double > &	Pi
	)

Constructor.

Initialises the necessary parameters and matrices

Parameters:

ns	number of dimensions in signal
nm	number of dimensions in measurement
nc	number of dimensions in control input
A	the ns*ns matrix relating signal at time k to signal at time k+1.
H	the nm*ns matrix relating the measurement to the signal.
B	the nc*ns matrix relating the signal to the control input. I'm not sure what this is used for, but I've included it for completeness.
x_initial	the initial ns*1 estimate of the signal.
p_initial	the initial error variance vector. P = p_initial.transpose()*p_initial.

Todo:: under development

Author:: Brendan McCane

Definition at line 54 of file kalman_filter.cxx.

KalmanFilter::KalmanFilter	(	unsigned int	ns,
		unsigned int	nm,
		const vnl_matrix< double > &	Ai,
		const vnl_matrix< double > &	Hi,
		const vnl_matrix< double > &	z_initial,
		const vnl_matrix< double > &	x_initial,
		const vnl_matrix< double > &	Pi
	)

Constructor.

Initialises the necessary parameters and matrices

Parameters:

ns	number of dimensions in signal
nm	number of dimensions in measurement
A	the ns*ns matrix relating signal at time k to signal at time k+1.
H	the nm*ns matrix relating the measurement to the signal.
x_initial	the initial ns*1 estimate of the signal.
p_initial	the initial error variance vector. P = p_initial.transpose()*p_initial.

Todo:: under development

Author:: Brendan McCane

Definition at line 131 of file kalman_filter.cxx.

Member Function Documentation

vnl_matrix<double> KalmanFilter::estimate ( ) const [inline]

Definition at line 117 of file kalman_filter.h.

void KalmanFilter::measurement_update	(	const vnl_matrix< double > &	zk,
		const vnl_matrix< double > &	Rk
	)

measurement_update.

Calculate the Kalman gain, update the signal estimate from the last time point using the previous estimate and the new measurement, and update the error covariance matrix.

Parameters:

zk	the new measurement
Rk	the measurement error covariance matrix

Todo:: under development

Author:: Brendan McCane

Definition at line 192 of file kalman_filter.cxx.

vnl_matrix< double > KalmanFilter::predict ( const vnl_matrix< double > & Qk )

predict.

Calculate the next state given the current estimate and project the error covariance matrix ahead in time as well. In this case, I am assuming there is no control input.

Parameters: