vpLinearKalmanFilterInstantiation.cpp

/****************************************************************************
 *
 * ViSP, open source Visual Servoing Platform software.
 * Copyright (C) 2005 - 2019 by Inria. All rights reserved.
 *
 * This software is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 * See the file LICENSE.txt at the root directory of this source
 * distribution for additional information about the GNU GPL.
 *
 * For using ViSP with software that can not be combined with the GNU
 * GPL, please contact Inria about acquiring a ViSP Professional
 * Edition License.
 *
 * See http://visp.inria.fr for more information.
 *
 * This software was developed at:
 * Inria Rennes - Bretagne Atlantique
 * Campus Universitaire de Beaulieu
 * 35042 Rennes Cedex
 * France
 *
 * If you have questions regarding the use of this file, please contact
 * Inria at visp@inria.fr
 *
 * This file is provided AS IS with NO WARRANTY OF ANY KIND, INCLUDING THE
 * WARRANTY OF DESIGN, MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
 *
 * Description:
 * Kalman filtering.
 *
 * Authors:
 * Eric Marchand
 * Fabien Spindler
 *
 *****************************************************************************/

#include <visp3/core/vpDebug.h>
#include <visp3/core/vpException.h>
#include <visp3/core/vpLinearKalmanFilterInstantiation.h>

#include <math.h>
#include <stdlib.h>

void vpLinearKalmanFilterInstantiation::initFilter(unsigned int n_signal, vpColVector &sigma_state,
                                                   vpColVector &sigma_measure, double rho, double delta_t)
{
  switch (model) {
  case stateConstVelWithColoredNoise_MeasureVel:
    initStateConstVelWithColoredNoise_MeasureVel(n_signal, sigma_state, sigma_measure, rho);
    break;
  case stateConstVel_MeasurePos:
    initStateConstVel_MeasurePos(n_signal, sigma_state, sigma_measure, delta_t);
    break;
  case stateConstAccWithColoredNoise_MeasureVel:
    initStateConstAccWithColoredNoise_MeasureVel(n_signal, sigma_state, sigma_measure, rho, delta_t);
    break;
  case unknown:
    vpERROR_TRACE("Kalman state model is not set");
    throw(vpException(vpException::notInitialized, "Kalman state model is not set"));
  }
}

void vpLinearKalmanFilterInstantiation::initStateConstVel_MeasurePos(unsigned int n_signal, vpColVector &sigma_state,
                                                                     vpColVector &sigma_measure, double delta_t)
{
  // init_done = true ;
  setStateModel(stateConstVel_MeasurePos);

  init(size_state, size_measure, n_signal);

  iter = 0;
  Pest = 0;
  Xest = 0;
  F = 0;
  H = 0;
  R = 0;
  Q = 0;
  this->dt = delta_t;

  double dt2 = dt * dt;
  double dt3 = dt2 * dt;

  for (unsigned int i = 0; i < size_measure * n_signal; i++) {
    // State model
    //         | 1  dt |
    //     F = |       |
    //         | 0   1 |

    F[2 * i][2 * i] = 1;
    F[2 * i][2 * i + 1] = dt;
    F[2 * i + 1][2 * i + 1] = 1;

    // Measure model
    H[i][2 * i] = 1;
    H[i][2 * i + 1] = 0;

    double sR = sigma_measure[i];
    double sQ = sigma_state[2 * i]; // sigma_state[2*i+1] is not used

    // Measure noise
    R[i][i] = sR;

    // State covariance matrix 6.2.2.12
    Q[2 * i][2 * i] = sQ * dt3 / 3;
    Q[2 * i][2 * i + 1] = sQ * dt2 / 2;
    Q[2 * i + 1][2 * i] = sQ * dt2 / 2;
    Q[2 * i + 1][2 * i + 1] = sQ * dt;

    Pest[2 * i][2 * i] = sR;
    Pest[2 * i][2 * i + 1] = sR / (2 * dt);
    Pest[2 * i + 1][2 * i] = sR / (2 * dt);
    Pest[2 * i + 1][2 * i + 1] = sQ * 2 * dt / 3.0 + sR / (2 * dt2);
  }
}

void vpLinearKalmanFilterInstantiation::initStateConstVelWithColoredNoise_MeasureVel(unsigned int n_signal,
                                                                                     vpColVector &sigma_state,
                                                                                     vpColVector &sigma_measure,
                                                                                     double rho)
{
  if ((rho < 0) || (rho >= 1)) {
    vpERROR_TRACE("Bad rho value %g; should be in [0:1[", rho);
    throw(vpException(vpException::badValue, "Bad rho value"));
  }

  setStateModel(stateConstVelWithColoredNoise_MeasureVel);

  init(size_state, size_measure, n_signal);

  iter = 0;
  Pest = 0;
  Xest = 0;
  F = 0;
  H = 0;
  R = 0;
  Q = 0;

  for (unsigned int i = 0; i < size_measure * n_signal; i++) {
    // State model
    //         | 1    1  |
    //     F = |         |
    //         | 0   rho |

    F[2 * i][2 * i] = 1;
    F[2 * i][2 * i + 1] = 1;
    F[2 * i + 1][2 * i + 1] = rho;

    // Measure model
    H[i][2 * i] = 1;
    H[i][2 * i + 1] = 0;

    double sR = sigma_measure[i];
    double sQ = sigma_state[2 * i + 1]; // sigma_state[2*i] is not used

    // Measure noise
    R[i][i] = sR;

    // State covariance matrix
    Q[2 * i][2 * i] = 0;
    Q[2 * i][2 * i + 1] = 0;
    Q[2 * i + 1][2 * i] = 0;
    Q[2 * i + 1][2 * i + 1] = sQ;

    Pest[2 * i][2 * i] = sR;
    Pest[2 * i][2 * i + 1] = 0.;
    Pest[2 * i + 1][2 * i] = 0;
    Pest[2 * i + 1][2 * i + 1] = sQ / (1 - rho * rho);
  }
}

void vpLinearKalmanFilterInstantiation::initStateConstAccWithColoredNoise_MeasureVel(unsigned int n_signal,
                                                                                     vpColVector &sigma_state,
                                                                                     vpColVector &sigma_measure,
                                                                                     double rho, double delta_t)
{
  if ((rho < 0) || (rho >= 1)) {
    vpERROR_TRACE("Bad rho value %g; should be in [0:1[", rho);
    throw(vpException(vpException::badValue, "Bad rho value"));
  }
  setStateModel(stateConstAccWithColoredNoise_MeasureVel);

  init(size_state, size_measure, n_signal);

  iter = 0;
  Pest = 0;
  Xest = 0;
  F = 0;
  H = 0;
  R = 0;
  Q = 0;
  this->dt = delta_t;
  // initialise les matrices decrivant les modeles
  for (unsigned int i = 0; i < size_measure * nsignal; i++) {
    // State model
    //         | 1    1   dt |
    //     F = | o   rho   0 |
    //         | 0    0    1 |

    F[3 * i][3 * i] = 1;
    F[3 * i][3 * i + 1] = 1;
    F[3 * i][3 * i + 2] = dt;
    F[3 * i + 1][3 * i + 1] = rho;
    F[3 * i + 2][3 * i + 2] = 1;

    // Measure model
    H[i][3 * i] = 1;
    H[i][3 * i + 1] = 0;
    H[i][3 * i + 2] = 0;

    double sR = sigma_measure[i];
    double sQ1 = sigma_state[3 * i + 1];
    double sQ2 = sigma_state[3 * i + 2];

    // Measure noise
    R[i][i] = sR;

    // State covariance matrix
    Q[3 * i + 1][3 * i + 1] = sQ1;
    Q[3 * i + 2][3 * i + 2] = sQ2;

    Pest[3 * i][3 * i] = sR;
    Pest[3 * i][3 * i + 1] = 0.;
    Pest[3 * i][3 * i + 2] = sR / dt;
    Pest[3 * i + 1][3 * i + 1] = sQ1 / (1 - rho * rho);
    Pest[3 * i + 1][3 * i + 2] = -rho * sQ1 / ((1 - rho * rho) * dt);
    Pest[3 * i + 2][3 * i + 2] = (2 * sR + sQ1 / (1 - rho * rho)) / (dt * dt);
    // complete the lower triangle
    Pest[3 * i + 1][3 * i] = Pest[3 * i][3 * i + 1];
    Pest[3 * i + 2][3 * i] = Pest[3 * i][3 * i + 2];
    Pest[3 * i + 2][3 * i + 1] = Pest[3 * i + 1][3 * i + 2];
  }
}

void vpLinearKalmanFilterInstantiation::filter(vpColVector &z)
{
  if (nsignal < 1) {
    vpERROR_TRACE("Bad signal number. You need to initialize the Kalman filter");
    throw(vpException(vpException::notInitialized, "Bad signal number"));
  }

  // Specific initialization of the filter that depends on the state model
  if (iter == 0) {
    Xest = 0;
    switch (model) {
    case stateConstVel_MeasurePos:
    case stateConstVelWithColoredNoise_MeasureVel:
    case stateConstAccWithColoredNoise_MeasureVel:
      for (unsigned int i = 0; i < size_measure * nsignal; i++) {
        Xest[size_state * i] = z[i];
      }
      prediction();
      //      init_done = true;
      break;
    case unknown:
      vpERROR_TRACE("Kalman state model is not set");
      throw(vpException(vpException::notInitialized, "Kalman state model is not set"));
      break;
    }
    iter++;

    return;
  } else if (iter == 1) {
    if (model == stateConstVel_MeasurePos) {
      for (unsigned int i = 0; i < size_measure * nsignal; i++) {
        double z_prev = Xest[size_state * i]; // Previous mesured position
        //  std::cout << "Mesure pre: " << z_prev << std::endl;
        Xest[size_state * i] = z[i];
        Xest[size_state * i + 1] = (z[i] - z_prev) / dt;
      }
      prediction();
      iter++;

      return;
    }
  }

  filtering(z);
  prediction();
}