javaslam.filter
Class LinearizedFilter

java.lang.Object
  |
  +--javaslam.filter.LinearizedFilter

All Implemented Interfaces:

NonlinearFilter

public class LinearizedFilter

extends Object

implements NonlinearFilter

A filter for nonlinear systems that uses local linearizations of the dynamics and measurement models so that a linear-Gaussian filter can be used to maintain an approximate Gaussian belief state. This class permits an arbitrary combination of linearization techniques and filter techniques; for example:

• by combining the UnscentedTransformation with the KalmanFilter we get the Unscented Kalman filter;
• by combining the ExtendedTransformation with the InformationFilter we get the Extended Information filter;

Field Summary

protected Filter	filter The underlying linear-Gaussian filter.
protected LinearizationFactory	linearization The factory for creating linearizations.

Constructor Summary

LinearizedFilter(Filter filter, LinearizationFactory linearization)
Constructor.

Method Summary

void	measurement(ListSet vars, NoisyVectorFunction g, double[] y) Performs an approximate nonlinear measurement update.
void	time(ListSet vars, NoisyVectorFunction f) Performs an approximate nonlinear time update.

Methods inherited from class java.lang.Object

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

Field Detail

filter

protected Filter filter

The underlying linear-Gaussian filter.

linearization

protected LinearizationFactory linearization

The factory for creating linearizations.

Constructor Detail

LinearizedFilter

public LinearizedFilter(Filter filter,
                        LinearizationFactory linearization)

Constructor.

Parameters:

filter - the underlying linear-Gaussian filter used for filtering

linearization - a linearization factory used to create linearizations of nonlinear models

Method Detail

measurement

public void measurement(ListSet vars,
                        NoisyVectorFunction g,
                        double[] y)

Performs an approximate nonlinear measurement update. The parameters vars, g, and q define the measurement equation as follows:

y=g(x[vars], v)

where v is an independent white noise vector. Given the actual measurement y, this method updates the belief state.

Specified by:

measurement in interface NonlinearFilter

Parameters:

vars - an ordered set of the variables with sum dimension n in the belief state that causally influenced this measurement.

g - the noisy measurement function with input dimension n + m and output dimension k; the input is formed from x[vars] stacked on top of the noise vector v

y - the measurement k-vector

Throws:

IllegalArgumentException - if there are any dimension mismatches

time

public void time(ListSet vars,
                 NoisyVectorFunction f)

Performs an approximate nonlinear time update. The parameters vars, f, and q define the state evolution equation as follows:

x_{t + 1}[vars]= f(x_t[vars], w)

where w is an independent white noise vector. All variables not in vars are assumed stationary.

Specified by:

time in interface NonlinearFilter

Parameters:

vars - an ordered set of the variables with sum dimension n in the belief state that evolve.

f - the state evolution function with input dimension n + m and output dimension n

Throws:

IllegalArgumentException - if there are any dimension mismatches