CRangeBearingKFSLAM.h

Go to the documentation of this file.

/* +------------------------------------------------------------------------+
   |                     Mobile Robot Programming Toolkit (MRPT)            |
   |                          http://www.mrpt.org/                          |
   |                                                                        |
   | Copyright (c) 2005-2018, 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 CRangeBearingKFSLAM_H
#define CRangeBearingKFSLAM_H

#include <mrpt/math/CMatrixTemplateNumeric.h>
#include <mrpt/config/CConfigFileBase.h>
#include <mrpt/config/CLoadableOptions.h>
#include <mrpt/bayes/CKalmanFilterCapable.h>

#include <mrpt/core/safe_pointers.h>
#include <mrpt/containers/bimap.h>

#include <mrpt/obs/CSensoryFrame.h>
#include <mrpt/obs/CActionCollection.h>
#include <mrpt/obs/CObservationBearingRange.h>
#include <mrpt/math/lightweight_geom_data.h>
#include <mrpt/poses/CPose3DPDFGaussian.h>
#include <mrpt/poses/CPose3DQuatPDFGaussian.h>
#include <mrpt/maps/CLandmark.h>
#include <mrpt/maps/CSimpleMap.h>
#include <mrpt/slam/CIncrementalMapPartitioner.h>
#include <mrpt/slam/data_association.h>

namespace mrpt::slam
{
/** An implementation of EKF-based SLAM with range-bearing sensors, odometry, a
 * full 6D robot pose, and 3D landmarks.
  *  The main method is "processActionObservation" which processes pairs of
 * action/observation.
  *  The state vector comprises: 3D robot position, a quaternion for its
 * attitude, and the 3D landmarks in the map.
  *
  *   The following Wiki page describes an front-end application based on this
 * class:
  *     http://www.mrpt.org/Application:kf-slam
  *
  *  For the theory behind this implementation, see the technical report in:
  *     http://www.mrpt.org/6D-SLAM
  *
  * \sa An implementation for 2D only: CRangeBearingKFSLAM2D
  * \ingroup metric_slam_grp
  */
class CRangeBearingKFSLAM
    : public bayes::CKalmanFilterCapable<7 /* x y z  qr qx qy qz*/,
                                         3 /* range yaw pitch */, 3 /* x y z */,
                                         7 /* Ax Ay Az Aqr Aqx Aqy Aqz */>
// <size_t VEH_SIZE,            size_t OBS_SIZE,        size_t FEAT_SIZE, size_t
// ACT_SIZE, size typename kftype = double>
{
   public:
    /** Either mrpt::math::TPoint2D or mrpt::math::TPoint3D */
    using landmark_point_t = mrpt::math::TPoint3D;

    /** Constructor. */
    CRangeBearingKFSLAM();

    /** Destructor: */
    virtual ~CRangeBearingKFSLAM();

    /** Reset the state of the SLAM filter: The map is emptied and the robot put
     * back to (0,0,0). */
    void reset();

    /** Process one new action and observations to update the map and robot pose
     *estimate. See the description of the class at the top of this page.
     *  \param action May contain odometry
     *  \param SF The set of observations, must contain at least one
     *CObservationBearingRange
     */
    void processActionObservation(
        mrpt::obs::CActionCollection::Ptr& action,
        mrpt::obs::CSensoryFrame::Ptr& SF);

    /** Returns the complete mean and cov.
      *  \param out_robotPose The mean and the 7x7 covariance matrix of the
     * robot 6D pose
      *  \param out_landmarksPositions One entry for each of the M landmark
     * positions (3D).
      *  \param out_landmarkIDs Each element[index] (for indices of
     * out_landmarksPositions) gives the corresponding landmark ID.
      *  \param out_fullState The complete state vector (7+3M).
      *  \param out_fullCovariance The full (7+3M)x(7+3M) covariance matrix of
     * the filter.
      * \sa getCurrentRobotPose
      */
    void getCurrentState(
        mrpt::poses::CPose3DQuatPDFGaussian& out_robotPose,
        std::vector<mrpt::math::TPoint3D>& out_landmarksPositions,
        std::map<unsigned int, mrpt::maps::CLandmark::TLandmarkID>&
            out_landmarkIDs,
        mrpt::math::CVectorDouble& out_fullState,
        mrpt::math::CMatrixDouble& out_fullCovariance) const;

    /** Returns the complete mean and cov.
      *  \param out_robotPose The mean and the 7x7 covariance matrix of the
     * robot 6D pose
      *  \param out_landmarksPositions One entry for each of the M landmark
     * positions (3D).
      *  \param out_landmarkIDs Each element[index] (for indices of
     * out_landmarksPositions) gives the corresponding landmark ID.
      *  \param out_fullState The complete state vector (7+3M).
      *  \param out_fullCovariance The full (7+3M)x(7+3M) covariance matrix of
     * the filter.
      * \sa getCurrentRobotPose
      */
    inline void getCurrentState(
        mrpt::poses::CPose3DPDFGaussian& out_robotPose,
        std::vector<mrpt::math::TPoint3D>& out_landmarksPositions,
        std::map<unsigned int, mrpt::maps::CLandmark::TLandmarkID>&
            out_landmarkIDs,
        mrpt::math::CVectorDouble& out_fullState,
        mrpt::math::CMatrixDouble& out_fullCovariance) const
    {
        mrpt::poses::CPose3DQuatPDFGaussian q(
            mrpt::math::UNINITIALIZED_QUATERNION);
        this->getCurrentState(
            q, out_landmarksPositions, out_landmarkIDs, out_fullState,
            out_fullCovariance);
        out_robotPose = mrpt::poses::CPose3DPDFGaussian(q);
    }

    /** Returns the mean & the 7x7 covariance matrix of the robot 6D pose (with
     * rotation as a quaternion).
      * \sa getCurrentState, getCurrentRobotPoseMean
      */
    void getCurrentRobotPose(
        mrpt::poses::CPose3DQuatPDFGaussian& out_robotPose) const;

    /** Get the current robot pose mean, as a 3D+quaternion pose.
      * \sa getCurrentRobotPose
      */
    mrpt::poses::CPose3DQuat getCurrentRobotPoseMean() const;

    /** Returns the mean & the 6x6 covariance matrix of the robot 6D pose (with
     * rotation as 3 angles).
      * \sa getCurrentState
      */
    inline void getCurrentRobotPose(
        mrpt::poses::CPose3DPDFGaussian& out_robotPose) const
    {
        mrpt::poses::CPose3DQuatPDFGaussian q(
            mrpt::math::UNINITIALIZED_QUATERNION);
        this->getCurrentRobotPose(q);
        out_robotPose = mrpt::poses::CPose3DPDFGaussian(q);
    }

    /** Returns a 3D representation of the landmarks in the map and the robot 3D
     * position according to the current filter state.
      *  \param out_objects
      */
    void getAs3DObject(mrpt::opengl::CSetOfObjects::Ptr& outObj) const;

    /** Load options from a ini-like file/text
      */
    void loadOptions(const mrpt::config::CConfigFileBase& ini);

    /** The options for the algorithm
      */
    struct TOptions : public mrpt::config::CLoadableOptions
    {
        /** Default values */
        TOptions();

        void loadFromConfigFile(
            const mrpt::config::CConfigFileBase& source,
            const std::string& section) override;  // See base docs
        void dumpToTextStream(
            std::ostream& out) const override;  // See base docs

        /** A 7-length vector with the std. deviation of the transition model in
         * (x,y,z, qr,qx,qy,qz) used only when there is no odometry (if there is
         * odo, its uncertainty values will be used instead); x y z: In meters.
         */
        mrpt::math::CVectorFloat stds_Q_no_odo;

        /** The std. deviation of the sensor (for the matrix R in the kalman
         * filters), in meters and radians. */
        float std_sensor_range, std_sensor_yaw, std_sensor_pitch;

        /** Additional std. dev. to sum to the motion model in the z axis
         * (useful when there is only 2D odometry and we want to put things hard
         * to the algorithm) (default=0) */
        float std_odo_z_additional;

        /** If set to true (default=false), map will be partitioned using the
         * method stated by partitioningMethod */
        bool doPartitioningExperiment;

        /** Default = 3 */
        float quantiles_3D_representation;

        /** Applicable only if "doPartitioningExperiment=true".
          *   0: Automatically detect partition through graph-cut.
          *   N>=1: Cut every "N" observations.
          */
        int partitioningMethod;

        // Data association:
        TDataAssociationMethod data_assoc_method;
        TDataAssociationMetric data_assoc_metric;
        /** Threshold in [0,1] for the chi2square test for individual
         * compatibility between predictions and observations (default: 0.99) */
        double data_assoc_IC_chi2_thres;
        /** Whether to use mahalanobis (->chi2 criterion) vs. Matching
         * likelihood. */
        TDataAssociationMetric data_assoc_IC_metric;
        /** Only if data_assoc_IC_metric==ML, the log-ML threshold (Default=0.0)
         */
        double data_assoc_IC_ml_threshold;

        /** Whether to fill m_SFs (default=false) */
        bool create_simplemap;

        /** Whether to ignore the input odometry and behave as if there was no
         * odometry at all (default: false) */
        bool force_ignore_odometry;
    } options;

    /** Information for data-association:
      * \sa getLastDataAssociation
      */
    struct TDataAssocInfo
    {
        TDataAssocInfo() : Y_pred_means(0, 0), Y_pred_covs(0, 0) {}
        void clear()
        {
            results.clear();
            predictions_IDs.clear();
            newly_inserted_landmarks.clear();
        }

        // Predictions from the map:
        mrpt::math::CMatrixTemplateNumeric<kftype> Y_pred_means, Y_pred_covs;
        std::vector<size_t> predictions_IDs;

        /** Map from the 0-based index within the last observation and the
           landmark 0-based index in the map (the robot-map state vector)
            Only used for stats and so. */
        std::map<size_t, size_t> newly_inserted_landmarks;

        // DA results:
        TDataAssociationResults results;
    };

    /** Returns a read-only reference to the information on the last
     * data-association */
    const TDataAssocInfo& getLastDataAssociation() const
    {
        return m_last_data_association;
    }

    /** Return the last partition of the sequence of sensoryframes (it is NOT a
     * partition of the map!!)
      *  Only if options.doPartitioningExperiment = true
      * \sa getLastPartitionLandmarks
      */
    void getLastPartition(std::vector<std::vector<uint32_t>>& parts)
    {
        parts = m_lastPartitionSet;
    }

    /** Return the partitioning of the landmarks in clusters accoring to the
     * last partition.
      *  Note that the same landmark may appear in different clusters (the
     * partition is not in the space of landmarks)
      *  Only if options.doPartitioningExperiment = true
      *  \param landmarksMembership The i'th element of this vector is the set
     * of clusters to which the i'th landmark in the map belongs to (landmark
     * index != landmark ID !!).
      * \sa getLastPartition
      */
    void getLastPartitionLandmarks(
        std::vector<std::vector<uint32_t>>& landmarksMembership) const;

    /** For testing only: returns the partitioning as
     * "getLastPartitionLandmarks" but as if a fixed-size submaps (size K) were
     * have been used.
      */
    void getLastPartitionLandmarksAsIfFixedSubmaps(
        size_t K, std::vector<std::vector<uint32_t>>& landmarksMembership);

    /** Computes the ratio of the missing information matrix elements which are
     * ignored under a certain partitioning of the landmarks.
      * \sa getLastPartitionLandmarks, getLastPartitionLandmarksAsIfFixedSubmaps
      */
    double computeOffDiagonalBlocksApproximationError(
        const std::vector<std::vector<uint32_t>>& landmarksMembership) const;

    /** The partitioning of the entire map is recomputed again.
      *  Only when options.doPartitioningExperiment = true.
      *  This can be used after changing the parameters of the partitioning
     * method.
      *  After this method, you can call getLastPartitionLandmarks.
      * \sa getLastPartitionLandmarks
      */
    void reconsiderPartitionsNow();

    /** Provides access to the parameters of the map partitioning algorithm.
      */
    CIncrementalMapPartitioner::TOptions* mapPartitionOptions()
    {
        return &mapPartitioner.options;
    }

    /** Save the current state of the filter (robot pose & map) to a MATLAB
     * script which displays all the elements in 2D
      */
    void saveMapAndPath2DRepresentationAsMATLABFile(
        const std::string& fil, float stdCount = 3.0f,
        const std::string& styleLandmarks = std::string("b"),
        const std::string& stylePath = std::string("r"),
        const std::string& styleRobot = std::string("r")) const;

   protected:
    /** @name Virtual methods for Kalman Filter implementation
        @{
     */

    /** Must return the action vector u.
      * \param out_u The action vector which will be passed to OnTransitionModel
      */
    void OnGetAction(KFArray_ACT& out_u) const;

    /** Implements the transition model \f$ \hat{x}_{k|k-1} = f(
     * \hat{x}_{k-1|k-1}, u_k ) \f$
      * \param in_u The vector returned by OnGetAction.
      * \param inout_x At input has \f[ \hat{x}_{k-1|k-1} \f] , at output must
     * have \f$ \hat{x}_{k|k-1} \f$ .
      * \param out_skip Set this to true if for some reason you want to skip the
     * prediction step (to do not modify either the vector or the covariance).
     * Default:false
      */
    void OnTransitionModel(
        const KFArray_ACT& in_u, KFArray_VEH& inout_x,
        bool& out_skipPrediction) const;

    /** Implements the transition Jacobian \f$ \frac{\partial f}{\partial x} \f$
      * \param out_F Must return the Jacobian.
      *  The returned matrix must be \f$V \times V\f$ with V being either the
     * size of the whole state vector (for non-SLAM problems) or VEH_SIZE (for
     * SLAM problems).
      */
    void OnTransitionJacobian(KFMatrix_VxV& out_F) const;

    /** Implements the transition noise covariance \f$ Q_k \f$
      * \param out_Q Must return the covariance matrix.
      *  The returned matrix must be of the same size than the jacobian from
     * OnTransitionJacobian
      */
    void OnTransitionNoise(KFMatrix_VxV& out_Q) const;

    /** This is called between the KF prediction step and the update step, and
     * the application must return the observations and, when applicable, the
     * data association between these observations and the current map.
      *
      * \param out_z N vectors, each for one "observation" of length OBS_SIZE, N
     * being the number of "observations": how many observed landmarks for a
     * map, or just one if not applicable.
      * \param out_data_association An empty vector or, where applicable, a
     * vector where the i'th element corresponds to the position of the
     * observation in the i'th row of out_z within the system state vector (in
     * the range [0,getNumberOfLandmarksInTheMap()-1]), or -1 if it is a new map
     * element and we want to insert it at the end of this KF iteration.
      * \param in_S The full covariance matrix of the observation predictions
     * (i.e. the "innovation covariance matrix"). This is a M*O x M*O matrix
     * with M=length of "in_lm_indices_in_S".
      * \param in_lm_indices_in_S The indices of the map landmarks (range
     * [0,getNumberOfLandmarksInTheMap()-1]) that can be found in the matrix
     * in_S.
      *
      *  This method will be called just once for each complete KF iteration.
      * \note It is assumed that the observations are independent, i.e. there
     * are NO cross-covariances between them.
      */
    void OnGetObservationsAndDataAssociation(
        vector_KFArray_OBS& out_z, std::vector<int>& out_data_association,
        const vector_KFArray_OBS& in_all_predictions, const KFMatrix& in_S,
        const std::vector<size_t>& in_lm_indices_in_S,
        const KFMatrix_OxO& in_R);

    void OnObservationModel(
        const std::vector<size_t>& idx_landmarks_to_predict,
        vector_KFArray_OBS& out_predictions) const;

    /** Implements the observation Jacobians \f$ \frac{\partial h_i}{\partial x}
     * \f$ and (when applicable) \f$ \frac{\partial h_i}{\partial y_i} \f$.
      * \param idx_landmark_to_predict The index of the landmark in the map
     * whose prediction is expected as output. For non SLAM-like problems, this
     * will be zero and the expected output is for the whole state vector.
      * \param Hx  The output Jacobian \f$ \frac{\partial h_i}{\partial x} \f$.
      * \param Hy  The output Jacobian \f$ \frac{\partial h_i}{\partial y_i}
     * \f$.
      */
    void OnObservationJacobians(
        const size_t& idx_landmark_to_predict, KFMatrix_OxV& Hx,
        KFMatrix_OxF& Hy) const;

    /** Computes A=A-B, which may need to be re-implemented depending on the
     * topology of the individual scalar components (eg, angles).
      */
    void OnSubstractObservationVectors(
        KFArray_OBS& A, const KFArray_OBS& B) const;

    /** Return the observation NOISE covariance matrix, that is, the model of
     * the Gaussian additive noise of the sensor.
      * \param out_R The noise covariance matrix. It might be non diagonal, but
     * it'll usually be.
      */
    void OnGetObservationNoise(KFMatrix_OxO& out_R) const;

    /** This will be called before OnGetObservationsAndDataAssociation to allow
     * the application to reduce the number of covariance landmark predictions
     * to be made.
      *  For example, features which are known to be "out of sight" shouldn't be
     * added to the output list to speed up the calculations.
      * \param in_all_prediction_means The mean of each landmark predictions;
     * the computation or not of the corresponding covariances is what we're
     * trying to determined with this method.
      * \param out_LM_indices_to_predict The list of landmark indices in the map
     * [0,getNumberOfLandmarksInTheMap()-1] that should be predicted.
      * \note This is not a pure virtual method, so it should be implemented
     * only if desired. The default implementation returns a vector with all the
     * landmarks in the map.
      * \sa OnGetObservations, OnDataAssociation
      */
    void OnPreComputingPredictions(
        const vector_KFArray_OBS& in_all_prediction_means,
        std::vector<size_t>& out_LM_indices_to_predict) const;

    /** If applicable to the given problem, this method implements the inverse
     * observation model needed to extend the "map" with a new "element".
      * \param in_z The observation vector whose inverse sensor model is to be
     * computed. This is actually one of the vector<> returned by
     * OnGetObservations().
      * \param out_yn The F-length vector with the inverse observation model \f$
     * y_n=y(x,z_n) \f$.
      * \param out_dyn_dxv The \f$F \times V\f$ Jacobian of the inv. sensor
     * model wrt the robot pose \f$ \frac{\partial y_n}{\partial x_v} \f$.
      * \param out_dyn_dhn The \f$F \times O\f$ Jacobian of the inv. sensor
     * model wrt the observation vector \f$ \frac{\partial y_n}{\partial h_n}
     * \f$.
      *
      *  - O: OBS_SIZE
      *  - V: VEH_SIZE
      *  - F: FEAT_SIZE
      *
      * \note OnNewLandmarkAddedToMap will be also called after calling this
     * method if a landmark is actually being added to the map.
      */
    void OnInverseObservationModel(
        const KFArray_OBS& in_z, KFArray_FEAT& out_yn,
        KFMatrix_FxV& out_dyn_dxv, KFMatrix_FxO& out_dyn_dhn) const;

    /** If applicable to the given problem, do here any special handling of
     * adding a new landmark to the map.
      * \param in_obsIndex The index of the observation whose inverse sensor is
     * to be computed. It corresponds to the row in in_z where the observation
     * can be found.
      * \param in_idxNewFeat The index that this new feature will have in the
     * state vector (0:just after the vehicle state, 1: after that,...). Save
     * this number so data association can be done according to these indices.
      * \sa OnInverseObservationModel
      */
    void OnNewLandmarkAddedToMap(
        const size_t in_obsIdx, const size_t in_idxNewFeat);

    /** This method is called after the prediction and after the update, to give
     * the user an opportunity to normalize the state vector (eg, keep angles
     * within -pi,pi range) if the application requires it.
      */
    void OnNormalizeStateVector();

    /** @}
     */

    /** Set up by processActionObservation */
    mrpt::obs::CActionCollection::Ptr m_action;

    /** Set up by processActionObservation */
    mrpt::obs::CSensoryFrame::Ptr m_SF;

    /** The mapping between landmark IDs and indexes in the Pkk cov. matrix: */
    mrpt::containers::bimap<mrpt::maps::CLandmark::TLandmarkID, unsigned int>
        m_IDs;

    /** Used for map partitioning experiments */
    CIncrementalMapPartitioner mapPartitioner;

    /** The sequence of all the observations and the robot path (kept for
     * debugging, statistics,etc)
      */
    mrpt::maps::CSimpleMap m_SFs;

    std::vector<std::vector<uint32_t>> m_lastPartitionSet;

    /** Last data association */
    TDataAssocInfo m_last_data_association;

    /** Return the last odometry, as a pose increment. */
    mrpt::poses::CPose3DQuat getIncrementFromOdometry() const;

};  // end class
}
#endif