Package org.hipparchus.filtering.kalman

Class KalmanSmoother

java.lang.Object

org.hipparchus.filtering.kalman.KalmanSmoother

All Implemented Interfaces:

KalmanObserver

public class KalmanSmoother
extends Object
implements KalmanObserver

Kalman smoother for linear, extended or unscented filters.

This implementation is attached to a filter using the observer mechanism. Once all measurements have been processed by the filter, the smoothing method can be called.

For example

     // Kalman filter
     final KalmanFilter<SimpleMeasurement> filter = new LinearKalmanFilter<>(decomposer, process, initialState);

     // Smoother observer
     final KalmanSmoother smoother = new KalmanSmoother(decomposer);
     filter.setObserver(smoother);

     // Process measurements with filter (forwards pass)
     measurements.forEach(filter::estimationStep);

     // Smooth backwards
     List<ProcessEstimate> smoothedStates = smoother.backwardsSmooth();
 

See Also:

• "Särkkä, S. Bayesian Filtering and Smoothing. Cambridge 2013"

Constructor Summary

Constructors

Constructor	Description
KalmanSmoother(MatrixDecomposer decomposer)	Simple constructor.

Method Summary

Modifier and Type	Method	Description
List<ProcessEstimate>	backwardsSmooth()	Backwards smooth.
void	init(KalmanEstimate estimate)	Callback for initialisation of observer.
void	updatePerformed(KalmanEstimate estimate)	Notification callback after each Kalman filter measurement update.

Methods inherited from class java.lang.Object

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

Constructor Details

KalmanSmoother

public KalmanSmoother(MatrixDecomposer decomposer)

Simple constructor.

Parameters:

decomposer - decomposer to use for the smoother gain calculations

Method Details

init

public void init(KalmanEstimate estimate)

Description copied from interface: KalmanObserver

Callback for initialisation of observer.

Specified by:

init in interface KalmanObserver

Parameters:

estimate - estimate calculated by a Kalman filter

updatePerformed

public void updatePerformed(KalmanEstimate estimate)

Description copied from interface: KalmanObserver

Notification callback after each Kalman filter measurement update.

Specified by:

updatePerformed in interface KalmanObserver

Parameters:

estimate - estimate calculated by a Kalman filter

backwardsSmooth

public List<ProcessEstimate> backwardsSmooth()

Backwards smooth. This is a backward pass over the filtered data, recursively calculating smoothed states, using the Rauch-Tung-Striebel (RTS) formulation. Note that the list result is a `LinkedList`, not an `ArrayList`.

Returns:

list of smoothed states