UnivariateProcessNoise.java
/* Copyright 2002-2019 CS Systèmes d'Information
* Licensed to CS Systèmes d'Information (CS) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* CS licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.orekit.estimation.sequential;
import org.hipparchus.analysis.UnivariateFunction;
import org.hipparchus.exception.LocalizedCoreFormats;
import org.hipparchus.linear.MatrixUtils;
import org.hipparchus.linear.RealMatrix;
import org.orekit.errors.OrekitException;
import org.orekit.frames.LOFType;
import org.orekit.orbits.PositionAngle;
import org.orekit.propagation.SpacecraftState;
import org.orekit.utils.CartesianDerivativesFilter;
/** Provider for a temporal evolution of the process noise matrix.
* All parameters (orbital or propagation) are time dependent and provided as {@link UnivariateFunction}.
* The argument of the functions is a duration in seconds (between current and previous spacecraft state).
* The output of the functions must be of the dimension of a standard deviation.
* The method {@link #getProcessNoiseMatrix} then square the values so that they are consistent with a covariance matrix.
* <p>
* The orbital parameters evolutions are provided in LOF frame and Cartesian (PV);
* then converted in inertial frame and current {@link org.orekit.orbits.OrbitType} and {@link PositionAngle}
* when method {@link #getProcessNoiseMatrix} is called.
* </p>
* <p>
* The time-dependent functions define a process noise matrix that is diagonal
* <em>in the Local Orbital Frame</em>, corresponds to Cartesian elements, abd represents
* the temporal evolution of (the standard deviation of) the process noise model. The
* first function is therefore the standard deviation along the LOF X axis, the second
* function represents the standard deviation along the LOF Y axis...
* This allows to set up simply a process noise representing an uncertainty that grows
* mainly along the track. The 6x6 upper left part of output matrix will however not be
* diagonal as it will be converted to the same inertial frame and orbit type as the
* {@link SpacecraftState state} used by the {@link KalmanEstimator Kalman estimator}.
* </p>
* <p>
* The propagation parameters are not associated to a specific frame and are appended as
* is in the lower right part diagonal of the output matrix. This implies this simplified
* model does not include correlation between the parameters and the orbit, but only
* evolution of the parameters themselves. If such correlations are needed, users must
* set up a custom {@link CovarianceMatrixProvider covariance matrix provider}. In most
* cases, the parameters are constant and their evolution noise is always 0, so the
* functions can be set to {@code x -> 0}.
* </p>
* <p>
* This class always provides one initial noise matrix or initial covariance matrix and one process noise matrix.
* </p>
* @author Luc Maisonobe
* @author Maxime Journot
* @since 9.2
*/
public class UnivariateProcessNoise extends AbstractCovarianceMatrixProvider {
/** Local Orbital Frame (LOF) type used. */
private final LOFType lofType;
/** Position angle for the orbital process noise matrix. */
private final PositionAngle positionAngle;
/** Array of univariate functions for the six orbital parameters process noise evolution in LOF frame and Cartesian formalism. */
private final UnivariateFunction[] lofCartesianOrbitalParametersEvolution;
/** Array of univariate functions for the propagation parameters process noise evolution. */
private final UnivariateFunction[] propagationParametersEvolution;
/** Simple constructor.
* @param initialCovarianceMatrix initial covariance matrix
* @param lofType the LOF type used
* @param positionAngle the position angle used for the computation of the process noise
* @param lofCartesianOrbitalParametersEvolution Array of univariate functions for the six orbital parameters process noise evolution in LOF frame and Cartesian orbit type
* @param propagationParametersEvolution Array of univariate functions for the propagation parameters process noise evolution
*/
public UnivariateProcessNoise(final RealMatrix initialCovarianceMatrix,
final LOFType lofType,
final PositionAngle positionAngle,
final UnivariateFunction[] lofCartesianOrbitalParametersEvolution,
final UnivariateFunction[] propagationParametersEvolution) {
super(initialCovarianceMatrix);
this.lofType = lofType;
this.positionAngle = positionAngle;
this.lofCartesianOrbitalParametersEvolution = lofCartesianOrbitalParametersEvolution.clone();
this.propagationParametersEvolution = propagationParametersEvolution.clone();
// Ensure that the orbital evolution array size is 6
if (lofCartesianOrbitalParametersEvolution.length != 6) {
throw new OrekitException(LocalizedCoreFormats.DIMENSIONS_MISMATCH,
lofCartesianOrbitalParametersEvolution, 6);
}
}
/** Getter for the lofType.
* @return the lofType
*/
public LOFType getLofType() {
return lofType;
}
/** Getter for the positionAngle.
* @return the positionAngle
*/
public PositionAngle getPositionAngle() {
return positionAngle;
}
/** Getter for the lofCartesianOrbitalParametersEvolution.
* @return the lofCartesianOrbitalParametersEvolution
*/
public UnivariateFunction[] getLofCartesianOrbitalParametersEvolution() {
return lofCartesianOrbitalParametersEvolution.clone();
}
/** Getter for the propagationParametersEvolution.
* @return the propagationParametersEvolution
*/
public UnivariateFunction[] getPropagationParametersEvolution() {
return propagationParametersEvolution.clone();
}
/** {@inheritDoc} */
@Override
public RealMatrix getProcessNoiseMatrix(final SpacecraftState previous,
final SpacecraftState current) {
// Orbital parameters process noise matrix in inertial frame and current orbit type
final RealMatrix inertialOrbitalProcessNoiseMatrix = getInertialOrbitalProcessNoiseMatrix(previous, current);
if (propagationParametersEvolution.length == 0) {
// no propagation parameters contribution, just return the orbital part
return inertialOrbitalProcessNoiseMatrix;
} else {
// Propagation parameters process noise matrix
final RealMatrix propagationProcessNoiseMatrix = getPropagationProcessNoiseMatrix(previous, current);
// Concatenate the matrices
final int orbitalMatrixSize = lofCartesianOrbitalParametersEvolution.length;
final int propagationMatrixSize = propagationParametersEvolution.length;
final RealMatrix processNoiseMatrix = MatrixUtils.createRealMatrix(orbitalMatrixSize + propagationMatrixSize,
orbitalMatrixSize + propagationMatrixSize);
processNoiseMatrix.setSubMatrix(inertialOrbitalProcessNoiseMatrix.getData(), 0, 0);
processNoiseMatrix.setSubMatrix(propagationProcessNoiseMatrix.getData(), orbitalMatrixSize, orbitalMatrixSize);
return processNoiseMatrix;
}
}
/** Get the process noise for the six orbital parameters in current spacecraft inertial frame and current orbit type.
* @param previous previous state
* @param current current state
* @return physical (i.e. non normalized) orbital process noise matrix in inertial frame
*/
private RealMatrix getInertialOrbitalProcessNoiseMatrix(final SpacecraftState previous,
final SpacecraftState current) {
// ΔT = duration in seconds from previous to current spacecraft state
final double deltaT = current.getDate().durationFrom(previous.getDate());
// Evaluate the functions, using ΔT as argument
final int lofOrbitalprocessNoiseLength = lofCartesianOrbitalParametersEvolution.length;
final double[] lofOrbitalProcessNoiseValues = new double[lofOrbitalprocessNoiseLength];
// The function return a value which dimension is that of a standard deviation
// It needs to be squared before being put in the process noise covariance matrix
for (int i = 0; i < lofOrbitalprocessNoiseLength; i++) {
final double functionValue = lofCartesianOrbitalParametersEvolution[i].value(deltaT);
lofOrbitalProcessNoiseValues[i] = functionValue * functionValue;
}
// Form the diagonal matrix in LOF frame and Cartesian formalism
final RealMatrix lofCartesianProcessNoiseMatrix = MatrixUtils.createRealDiagonalMatrix(lofOrbitalProcessNoiseValues);
// Get the Jacobian from LOF to Inertial
final double[][] dLofdInertial = new double[6][6];
lofType.transformFromInertial(current.getDate(),
current.getPVCoordinates()).getInverse().getJacobian(CartesianDerivativesFilter.USE_PV,
dLofdInertial);
final RealMatrix jacLofToInertial = MatrixUtils.createRealMatrix(dLofdInertial);
// Jacobian of orbit parameters with respect to Cartesian parameters
final double[][] dYdC = new double[6][6];
current.getOrbit().getJacobianWrtCartesian(positionAngle, dYdC);
final RealMatrix jacOrbitWrtCartesian = MatrixUtils.createRealMatrix(dYdC);
// Complete Jacobian of the transformation
final RealMatrix jacobian = jacOrbitWrtCartesian.multiply(jacLofToInertial);
// Return the orbital process noise matrix in inertial frame and proper orbit type
final RealMatrix inertialOrbitalProcessNoiseMatrix = jacobian.
multiply(lofCartesianProcessNoiseMatrix).
multiplyTransposed(jacobian);
return inertialOrbitalProcessNoiseMatrix;
}
/** Get the process noise for the propagation parameters.
* @param previous previous state
* @param current current state
* @return physical (i.e. non normalized) propagation process noise matrix
*/
private RealMatrix getPropagationProcessNoiseMatrix(final SpacecraftState previous,
final SpacecraftState current) {
// ΔT = duration from previous to current spacecraft state (in seconds)
final double deltaT = current.getDate().durationFrom(previous.getDate());
// Evaluate the functions, using ΔT as argument
final int propagationProcessNoiseLength = propagationParametersEvolution.length;
final double[] propagationProcessNoiseValues = new double[propagationProcessNoiseLength];
// The function return a value which dimension is that of a standard deviation
// It needs to be squared before being put in the process noise covariance matrix
for (int i = 0; i < propagationProcessNoiseLength; i++) {
final double functionValue = propagationParametersEvolution[i].value(deltaT);
propagationProcessNoiseValues[i] = functionValue * functionValue;
}
// Form the diagonal matrix corresponding to propagation parameters process noise
return MatrixUtils.createRealDiagonalMatrix(propagationProcessNoiseValues);
}
}