CPointPDFSOG.h

Go to the documentation of this file.

/* +------------------------------------------------------------------------+
   |                     Mobile Robot Programming Toolkit (MRPT)            |
   |                          https://www.mrpt.org/                         |
   |                                                                        |
   | Copyright (c) 2005-2019, Individual contributors, see AUTHORS file     |
   | See: https://www.mrpt.org/Authors - All rights reserved.               |
   | Released under BSD License. See: https://www.mrpt.org/License          |
   +------------------------------------------------------------------------+ */
#pragma once

#include <mrpt/math/CMatrixD.h>
#include <mrpt/math/CMatrixF.h>
#include <mrpt/poses/CPointPDF.h>
#include <mrpt/poses/CPointPDFGaussian.h>

namespace mrpt::poses
{
/** Declares a class that represents a Probability Density function (PDF) of a
 * 3D point \f$ p(\mathbf{x}) = [x ~ y ~ z ]^t \f$.
 *   This class implements that PDF as the following multi-modal Gaussian
 * distribution:
 *
 * \f$ p(\mathbf{x}) = \sum\limits_{i=1}^N \omega^i \mathcal{N}( \mathbf{x} ;
 * \bar{\mathbf{x}}^i, \mathbf{\Sigma}^i )  \f$
 *
 *  Where the number of modes N is the size of CPointPDFSOG::m_modes
 *
 *  See mrpt::poses::CPointPDF for more details.
 *
 * \sa CPointPDF, CPosePDF,
 * \ingroup poses_pdf_grp
 */
class CPointPDFSOG : public CPointPDF
{
    DEFINE_SERIALIZABLE(CPointPDFSOG)

   public:
    /** The struct for each mode:
     */
    struct TGaussianMode
    {
        TGaussianMode() : val() {}
        CPointPDFGaussian val;

        /** The log-weight
         */
        double log_w{0};
    };

    using CListGaussianModes = std::deque<TGaussianMode>;
    using const_iterator = std::deque<TGaussianMode>::const_iterator;
    using iterator = std::deque<TGaussianMode>::iterator;

   protected:
    /** Assures the symmetry of the covariance matrix (eventually certain
     * operations in the math-coprocessor lead to non-symmetric matrixes!)
     */
    void enforceCovSymmetry();

    /** The list of SOG modes */
    CListGaussianModes m_modes;

   public:
    /** Default constructor
     * \param nModes The initial size of CPointPDFSOG::m_modes
     */
    CPointPDFSOG(size_t nModes = 1);

    /** Clear all the gaussian modes */
    void clear();

    /** Access to individual beacons */
    const TGaussianMode& operator[](size_t i) const
    {
        ASSERT_(i < m_modes.size());
        return m_modes[i];
    }
    /** Access to individual beacons */
    TGaussianMode& operator[](size_t i)
    {
        ASSERT_(i < m_modes.size());
        return m_modes[i];
    }

    /** Access to individual beacons */
    const TGaussianMode& get(size_t i) const
    {
        ASSERT_(i < m_modes.size());
        return m_modes[i];
    }
    /** Access to individual beacons */
    TGaussianMode& get(size_t i)
    {
        ASSERT_(i < m_modes.size());
        return m_modes[i];
    }

    /** Inserts a copy of the given mode into the SOG */
    void push_back(const TGaussianMode& m) { m_modes.push_back(m); }
    iterator begin() { return m_modes.begin(); }
    iterator end() { return m_modes.end(); }
    const_iterator begin() const { return m_modes.begin(); }
    const_iterator end() const { return m_modes.end(); }
    iterator erase(iterator i) { return m_modes.erase(i); }
    /** Resize the number of SOG modes */
    void resize(const size_t N);
    /** Return the number of Gaussian modes. */
    size_t size() const { return m_modes.size(); }
    /** Return whether there is any Gaussian mode. */
    bool empty() const { return m_modes.empty(); }

    void getMean(CPoint3D& mean_point) const override;

    std::tuple<cov_mat_t, type_value> getCovarianceAndMean() const override;

    /** Normalize the weights in m_modes such as the maximum log-weight is 0 */
    void normalizeWeights();

    /** Return the Gaussian mode with the highest likelihood (or an empty
     * Gaussian if there are no modes in this SOG) */
    void getMostLikelyMode(CPointPDFGaussian& outVal) const;

    /** Computes the "Effective sample size" (typical measure for Particle
     * Filters), applied to the weights of the individual Gaussian modes, as a
     * measure of the equality of the modes (in the range [0,total # of modes]).
     */
    double ESS() const;

    /** Copy operator, translating if necesary (for example, between particles
     * and gaussian representations) */
    void copyFrom(const CPointPDF& o) override;

    /** Save the density to a text file, with the following format:
     *  There is one row per Gaussian "mode", and each row contains 10
     * elements:
     *   - w (The weight)
     *   - x_mean (gaussian mean value)
     *   - y_mean (gaussian mean value)
     *   - x_mean (gaussian mean value)
     *   - C11 (Covariance elements)
     *   - C22 (Covariance elements)
     *   - C33 (Covariance elements)
     *   - C12 (Covariance elements)
     *   - C13 (Covariance elements)
     *   - C23 (Covariance elements)
     *
     */
    bool saveToTextFile(const std::string& file) const override;

    /** this = p (+) this. This can be used to convert a PDF from local
     * coordinates to global, providing the point (newReferenceBase) from which
     *   "to project" the current pdf. Result PDF substituted the currently
     * stored one in the object. */
    void changeCoordinatesReference(const CPose3D& newReferenceBase) override;

    /** Draw a sample from the pdf. */
    void drawSingleSample(CPoint3D& outSample) const override;

    /** Bayesian fusion of two point distributions (product of two
     * distributions->new distribution), then save the result in this object
     * (WARNING: See implementing classes to see classes that can and cannot be
     * mixtured!)
     * \param p1 The first distribution to fuse
     * \param p2 The second distribution to fuse
     * \param minMahalanobisDistToDrop If set to different of 0, the result of
     * very separate Gaussian modes (that will result in negligible components)
     * in SOGs will be dropped to reduce the number of modes in the output.
     */
    void bayesianFusion(
        const CPointPDF& p1, const CPointPDF& p2,
        const double minMahalanobisDistToDrop = 0) override;

    /** Evaluates the PDF within a rectangular grid and saves the result in a
     * matrix (each row contains values for a fixed y-coordinate value).
     */
    void evaluatePDFInArea(
        float x_min, float x_max, float y_min, float y_max, float resolutionXY,
        float z, mrpt::math::CMatrixD& outMatrix, bool sumOverAllZs = false);

    /** Evaluates the PDF at a given point */
    double evaluatePDF(const CPoint3D& x, bool sumOverAllZs) const;

};  // End of class def.
}  // namespace mrpt::poses