Package org.hipparchus.filtering.kalman

Class AbstractKalmanFilter<T extends Measurement>

java.lang.Object

org.hipparchus.filtering.kalman.AbstractKalmanFilter<T>

Type Parameters:

T - the type of the measurements

All Implemented Interfaces:

KalmanFilter<T>

Direct Known Subclasses:

ExtendedKalmanFilter, LinearKalmanFilter

public abstract class AbstractKalmanFilter<T extends Measurement>
extends Object
implements KalmanFilter<T>

Shared parts between linear and non-linear Kalman filters.

Since:

1.3

Constructor Summary

Constructors

Modifier	Constructor	Description
protected	AbstractKalmanFilter(MatrixDecomposer decomposer, ProcessEstimate initialState)	Simple constructor.

Method Summary

Modifier and Type	Method	Description
protected RealMatrix	computeInnovationCovarianceMatrix(RealMatrix r, RealMatrix h)	Compute innovation covariance matrix.
protected void	correct(T measurement, RealVector innovation, RealMatrix h, RealMatrix s)	Perform correction step.
ProcessEstimate	getCorrected()	Get the corrected state.
ProcessEstimate	getPredicted()	Get the predicted state.
protected void	predict(double time, RealVector predictedState, RealMatrix stm, RealMatrix noise)	Perform prediction step.

Methods inherited from interface org.hipparchus.filtering.kalman.KalmanFilter

estimationStep

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

AbstractKalmanFilter

protected AbstractKalmanFilter(MatrixDecomposer decomposer,
                               ProcessEstimate initialState)

Simple constructor.

Parameters:

decomposer - decomposer to use for the correction phase

initialState - initial state

Method Detail

predict

protected void predict(double time,
                       RealVector predictedState,
                       RealMatrix stm,
                       RealMatrix noise)

Perform prediction step.

Parameters:

time - process time

predictedState - predicted state vector

stm - state transition matrix

noise - process noise covariance matrix

computeInnovationCovarianceMatrix

protected RealMatrix computeInnovationCovarianceMatrix(RealMatrix r,
                                                       RealMatrix h)

Compute innovation covariance matrix.

Parameters:

r - measurement covariance

h - Jacobian of the measurement with respect to the state (may be null if measurement should be ignored)

Returns:

innovation covariance matrix, defined as \(h.P.h^T + r\), or null if h is null

correct

protected void correct(T measurement,
                       RealVector innovation,
                       RealMatrix h,
                       RealMatrix s)
                throws MathIllegalArgumentException

Perform correction step.

Parameters:

measurement - single measurement to handle

innovation - innovation vector (i.e. residuals) (may be null if measurement should be ignored)

h - Jacobian of the measurement with respect to the state (may be null if measurement should be ignored)

s - innovation covariance matrix (may be null if measurement should be ignored)

Throws:

MathIllegalArgumentException - if matrix cannot be decomposed

getPredicted

public ProcessEstimate getPredicted()

Get the predicted state.

Specified by:

getPredicted in interface KalmanFilter<T extends Measurement>

Returns:

predicted state

getCorrected

public ProcessEstimate getCorrected()

Get the corrected state.

Specified by:

getCorrected in interface KalmanFilter<T extends Measurement>

Returns:

corrected state