CMetricMapBuilderRBPF.h

Go to the documentation of this file.

/* +------------------------------------------------------------------------+
   |                     Mobile Robot Programming Toolkit (MRPT)            |
   |                          http://www.mrpt.org/                          |
   |                                                                        |
   | Copyright (c) 2005-2017, Individual contributors, see AUTHORS file     |
   | See: http://www.mrpt.org/Authors - All rights reserved.                |
   | Released under BSD License. See details in http://www.mrpt.org/License |
   +------------------------------------------------------------------------+ */
#ifndef CMetricMapBuilderRBPF_H
#define CMetricMapBuilderRBPF_H

#include <mrpt/slam/CMetricMapBuilder.h>
#include <mrpt/maps/CMultiMetricMapPDF.h>
#include <mrpt/maps/CMultiMetricMap.h>

#include <mrpt/bayes/CParticleFilter.h>
#include <mrpt/utils/CLoadableOptions.h>
#include <mrpt/utils/safe_pointers.h>

namespace mrpt
{
namespace slam
{
/** This class implements a Rao-Blackwelized Particle Filter (RBPF) approach to
 * map building (SLAM).
 *   Internally, the list of particles, each containing a hypothesis for the
 * robot path plus its associated
 *    metric map, is stored in an object of class CMultiMetricMapPDF.
 *
 *  This class processes robot actions and observations sequentially (through
 * the method CMetricMapBuilderRBPF::processActionObservation)
 *   and exploits the generic design of metric map classes in MRPT to deal with
 * any number and combination of maps simultaneously: the likelihood
 *   of observations is the product of the likelihood in the different maps,
 * etc.
 *
 *   A number of particle filter methods are implemented as well, by selecting
 * the appropriate values in TConstructionOptions::PF_options.
 *   Not all the PF algorithms are implemented for all kinds of maps.
 *
 *  For an example of usage, check the application "rbpf-slam", in
 * "apps/RBPF-SLAM". See also the <a
 * href="http://www.mrpt.org/Application:RBPF-SLAM" >wiki page</a>.
 *
 *  \note Since MRPT 0.7.2, the new variables
 * "localizeLinDistance,localizeAngDistance" are introduced to provide a way to
 * update the robot pose at a different rate than the map is updated.
 *  \note Since MRPT 0.7.1 the semantics of the parameters
 * "insertionLinDistance" and "insertionAngDistance" changes: the entire RBFP is
 * now NOT updated unless odometry increments surpass the threshold (previously,
 * only the map was NOT updated). This is done to gain efficiency.
 *  \note Since MRPT 0.6.2 this class implements full 6D SLAM. Previous versions
 * worked in 2D + heading only.
 *
 * \sa CMetricMap   \ingroup metric_slam_grp
 */
class CMetricMapBuilderRBPF : public mrpt::slam::CMetricMapBuilder
{
   public:
        /** The map PDF: It includes a path and associated map for each particle. */
        mrpt::maps::CMultiMetricMapPDF mapPDF;

   protected:
        /** The configuration of the particle filter */
        bayes::CParticleFilter::TParticleFilterOptions m_PF_options;

        /** Distances (linear and angular) for inserting a new observation into the
         * map. */
        float insertionLinDistance, insertionAngDistance;

        /** Distances (linear and angular) for updating the robot pose estimate (and
         * particles weighs, if applicable). */
        float localizeLinDistance, localizeAngDistance;

        /** Traveled distance since last localization update */
        mrpt::poses::CPose3DPDFGaussian odoIncrementSinceLastLocalization;
        /** Traveled distance since last map update */
        mrpt::poses::CPose3D odoIncrementSinceLastMapUpdate;

   public:
        /** Options for building a CMetricMapBuilderRBPF object, passed to the
         * constructor.
         */
        struct TConstructionOptions : public utils::CLoadableOptions
        {
           public:
                /** Constructor */
                TConstructionOptions();
                void loadFromConfigFile(
                        const mrpt::utils::CConfigFileBase& source,
                        const std::string& section) override;  // See base docs
                void dumpToTextStream(
                        mrpt::utils::CStream& out) const override;  // See base docs

                float insertionLinDistance;
                float insertionAngDistance;

                float localizeLinDistance;
                float localizeAngDistance;

                bayes::CParticleFilter::TParticleFilterOptions PF_options;

                mrpt::maps::TSetOfMetricMapInitializers mapsInitializers;
                mrpt::maps::CMultiMetricMapPDF::TPredictionParams predictionOptions;
                mrpt::utils::VerbosityLevel verbosity_level;
        };

        /** Constructor. */
        CMetricMapBuilderRBPF(const TConstructionOptions& initializationOptions);

        /** This second constructor is created for the situation where a class
member needs to be
of type CMetricMapBuilderRBPF  */
        CMetricMapBuilderRBPF();

        /** Copy Operator. */
        CMetricMapBuilderRBPF& operator=(const CMetricMapBuilderRBPF& src);

        /** Destructor. */
        virtual ~CMetricMapBuilderRBPF();

        /** Initialize the method, starting with a known location PDF "x0"(if
         * supplied, set to nullptr to left unmodified) and a given fixed, past map.
         */
        void initialize(
                const mrpt::maps::CSimpleMap& initialMap = mrpt::maps::CSimpleMap(),
                const mrpt::poses::CPosePDF* x0 = nullptr);

        /** Clear all elements of the maps.
         */
        void clear();

        /** Returns a copy of the current best pose estimation as a pose PDF.
         */
        mrpt::poses::CPose3DPDF::Ptr getCurrentPoseEstimation() const;

        /** Returns the current most-likely path estimation (the path associated to
         * the most likely particle).
         */
        void getCurrentMostLikelyPath(
                std::deque<mrpt::math::TPose3D>& outPath) const;

        /** Appends a new action and observations to update this map: See the
         *description of the class at the top of this page to see a more complete
         *description.
         *  \param action The incremental 2D pose change in the robot pose. This
         *value is deterministic.
         *      \param observations The set of observations that robot senses at the new
         *pose.
         *  Statistics will be saved to statsLastIteration
         */
        void processActionObservation(
                mrpt::obs::CActionCollection& action,
                mrpt::obs::CSensoryFrame& observations);

        /** Fills "out_map" with the set of "poses"-"sensory-frames", thus the so
         * far built map.
         */
        void getCurrentlyBuiltMap(mrpt::maps::CSimpleMap& out_map) const;

        /** Returns the map built so far. NOTE that for efficiency a pointer to the
         * internal object is passed, DO NOT delete nor modify the object in any
         * way, if desired, make a copy of ir with "clone()".
         */
        const mrpt::maps::CMultiMetricMap* getCurrentlyBuiltMetricMap() const;

        /** Returns just how many sensory-frames are stored in the currently build
         * map.
         */
        unsigned int getCurrentlyBuiltMapSize();

        /** A useful method for debugging: the current map (and/or poses) estimation
         * is dumped to an image file.
         * \param file The output file name
         * \param formatEMF_BMP Output format = true:EMF, false:BMP
         */
        void saveCurrentEstimationToImage(
                const std::string& file, bool formatEMF_BMP = true);

        /** A useful method for debugging: draws the current map and path hypotheses
         * to a CCanvas  */
        void drawCurrentEstimationToImage(utils::CCanvas* img);

        /** A logging utility: saves the current path estimation for each particle
         * in a text file (a row per particle, each 3-column-entry is a set
         * [x,y,phi], respectively).
         */
        void saveCurrentPathEstimationToTextFile(const std::string& fil);

        double getCurrentJointEntropy();

        /** This structure will hold stats after each execution of
         * processActionObservation
         */
        struct TStats
        {
                TStats() : observationsInserted(false) {}
                /** Whether the SF has been inserted in the metric maps. */
                bool observationsInserted;
        };

        /** This structure will hold stats after each execution of
         * processActionObservation */
        TStats m_statsLastIteration;

};  // End of class def.

}  // namespace slam
}  // namespace mrpt
#endif