[mrpt-obs]

Overview

Observation classes for all kinds of robot sensors.

Library mrpt-obs

This library is part of MRPT and can be installed in Debian-based systems with:

sudo apt install libmrpt-obs-dev

Key concepts:

Sensor observations : All sensor observations share a common virtual base class (mrpt::obs::CObservation). There are classes to store 2D/3D laser scans, rotating LiDAR scans, monocular and stereo images, depth cameras, GPS, IMU, odometry, gas sensors, WiFi/RFID, and more. A concept closely related to observations is mrpt::obs::CSensoryFrame — a set of observations collected approximately at the same instant.
Rawlogs (datasets) : A robotics dataset can be loaded, edited and explored by means of mrpt::obs::CRawlog. See also: https://www.mrpt.org/Rawlog_Format
Actions : For Bayesian filtering, robot actions (e.g. 2D odometry increments) can be represented as mrpt::obs::CAction objects.
Simple maps : A CSimpleMap is a sequence of (pose PDF, sensory frame) pairs. Metric maps can be rebuilt on-demand from this compact representation.
CARMEN log tools : Utilities to convert CARMEN log files to MRPT rawlogs. See mrpt::obs::carmen_log_parse_line and the carmen2rawlog / carmen2simplemap applications.

Key observation classes (MRPT 3 API)

Class	Sensor
mrpt::obs::CObservation2DRangeScan	2D laser scanner (single scan arc)
mrpt::obs::CObservationRotatingScan	Rotating LiDAR (organized range/intensity images); replaces `CObservationVelodyneScan` for processing
mrpt::obs::CObservationVelodyneScan	Raw Velodyne packets; convert via `CObservationRotatingScan::fromVelodyne()`
mrpt::obs::CObservationPointCloud	Unstructured 3D point cloud
mrpt::obs::CObservationImage	Monocular camera image
mrpt::obs::CObservationStereoImages	Stereo pair
mrpt::obs::CObservation3DRangeScan	RGB-D (structured depth + color)
mrpt::obs::CObservationGPS	GPS/GNSS fix
mrpt::obs::CObservationIMU	IMU (accelerometer, gyro, magnetometer)
mrpt::obs::CObservationOdometry	Wheel odometry
mrpt::obs::CObservationGasSensors	Electronic nose / gas sensors
mrpt::obs::CObservationWirelessPower	WiFi RSSI

Library contents

// enums

enum mrpt::obs::GnssFixType;
enum mrpt::obs::GnssService;
enum mrpt::obs::TIMUDataIndex;

// structs

struct mrpt::obs::T2DScanProperties;
struct mrpt::obs::T3DPointsProjectionParams;
struct mrpt::obs::T3DPointsTo2DScanParams;
struct mrpt::maps::TMetricMapInitializer;

template <unsigned int BYTES_REQUIRED_>
struct mrpt::obs::TPixelLabelInfo;

struct mrpt::obs::TPixelLabelInfoBase;
struct mrpt::obs::VelodyneCalibration;

// classes

class mrpt::obs::CAction;
class mrpt::obs::CActionCollection;
class mrpt::obs::CActionRobotMovement2D;
class mrpt::obs::CActionRobotMovement3D;
class mrpt::maps::CMetricMap;
class mrpt::obs::CObservation2DRangeScan;
class mrpt::obs::CObservation3DRangeScan;
class mrpt::obs::CObservation3DScene;
class mrpt::obs::CObservation6DFeatures;
class mrpt::obs::CObservationBatteryState;
class mrpt::obs::CObservationBeaconRanges;
class mrpt::obs::CObservationBearingRange;
class mrpt::obs::CObservationCANBusJ1939;
class mrpt::obs::CObservationComment;
class mrpt::obs::CObservationGPS;
class mrpt::obs::CObservationGasSensors;
class mrpt::obs::CObservationIMU;
class mrpt::obs::CObservationImage;
class mrpt::obs::CObservationOdometry;
class mrpt::obs::CObservationRFID;
class mrpt::obs::CObservationRGBD360;
class mrpt::obs::CObservationRange;
class mrpt::obs::CObservationRawDAQ;
class mrpt::obs::CObservationReflectivity;
class mrpt::obs::CObservationRobotPose;
class mrpt::obs::CObservationRotatingScan;
class mrpt::obs::CObservationSkeleton;
class mrpt::obs::CObservationStereoImages;
class mrpt::obs::CObservationStereoImagesFeatures;
class mrpt::obs::CObservationVelodyneScan;
class mrpt::obs::CObservationWindSensor;
class mrpt::obs::CObservationWirelessPower;
class mrpt::obs::CRawlog;
class mrpt::obs::CSensoryFrame;
class mrpt::maps::CSimpleMap;
class mrpt::obs::CSinCosLookUpTableFor2DScans;
class mrpt::maps::TSetOfMetricMapInitializers;
class mrpt::maps::mrptEventMetricMapClear;
class mrpt::maps::mrptEventMetricMapInsert;

// global variables

static constexpr int mrpt::obs::INVALID_BEACON_ID = -1;
static constexpr int mrpt::obs::INVALID_LANDMARK_ID = -1;

// global functions

bool mrpt::obs::carmen_log_parse_line(
    std::istream& in_stream,
    std::vector<mrpt::obs::CObservation::Ptr>& out_imported_observations,
    const mrpt::system::TTimeStamp& time_start_log
    );

std::string mrpt::obs::format_externals_filename(const mrpt::obs::CObservation& obs, const std::string& fmt);
void mrpt::obs::registerAllClasses_mrpt_obs();
void mrpt::obs::stock_observations::example2DRangeScan(mrpt::obs::CObservation2DRangeScan& s, int i = 0);
void mrpt::obs::stock_observations::exampleImage(mrpt::img::CImage& im, int i = 0);

template <class OBSERVATION_T>
OBSERVATION_T::Ptr mrpt::obs::utils::getObservation(
    mrpt::obs::CSensoryFrame::Ptr& observations,
    mrpt::obs::CObservation::Ptr& observation,
    bool priority_to_sf = true
    );

Global Variables

static constexpr int mrpt::obs::INVALID_BEACON_ID = -1

Used for CObservationBeaconRange, CBeacon, etc.

static constexpr int mrpt::obs::INVALID_LANDMARK_ID = -1

Used for CObservationBearingRange::TMeasurement::beaconID and others.

Global Functions

bool mrpt::obs::carmen_log_parse_line(
    std::istream& in_stream,
    std::vector<mrpt::obs::CObservation::Ptr>& out_imported_observations,
    const mrpt::system::TTimeStamp& time_start_log
    )

Parse one line from an text input stream and interpret it as a CARMEN log entry, returning its MRPT observation representation.

The first word in each line determines the type of that entry. Supported line entries in this class are the following:

“ROBOTLASER(.*)”: A joint odometry-laser entry. This function will return both an mrpt::obs::CObservationOdometry and an mrpt::obs::CObservation2DRangeScan
“FLASER”: (Idem)
“RLASER”: (Idem)
“PARAM”: This case is special (read below).

Note that the sensor position on the robot will be always deduced from the corresponding CARMEN log line automatically.

The following entry types will be IGNORED, since with the tags above will be enough in most applications.

“ODOM”
“RAWLASER”

About the PARAM entries: This functions has an internal static status consisting on some PARAM values which have effects on the future read observations. When the function finds one of these, it stores the parameter value, return an empty vector of observations, and use those values for future laser entries of types FLASER or RLASER. Currently, only these parameters are processed:

“robot_front_laser_max”
“laser_front_laser_resolution”
“robot_rear_laser_max”
“laser_rear_laser_resolution”

References: http://carmen.sourceforge.net/doc/binary__loggerplayback.html

Parameters:

time_start_log	The real timestamp that corresponds to a “0” in the CARMEN log (you can get a mrpt::Clock::now() once and pass that same value with each call).
std::runtime_error	On any invalid line found.

Returns:

true on success, false on end-of-file (EOF).

std::string mrpt::obs::format_externals_filename(
    const mrpt::obs::CObservation& obs,
    const std::string& fmt
    )

Replaces format placeholders in a string according to an observation:

${type} is replaced by: img, stereo, 3dcam for single images, stereo images and depth camera observations, respectively, or other otherwise.
${label} is replaced by the observation sensorLabel field.
f with any standard printf() format modifiers will be replaced by UNIX timestamp (with fractions of seconds) of the observation.
Anything else will be left unmodified.

For example, the default format string used in rawlog-edit --rename-externals is "${type}_${label}_%.06f".

(new in MRPT 2.4.1)

void mrpt::obs::registerAllClasses_mrpt_obs()

Forces manual RTTI registration of all serializable classes in this namespace.

Should never be required to be explicitly called by users, except if building MRPT as a static library.

void mrpt::obs::stock_observations::example2DRangeScan(mrpt::obs::CObservation2DRangeScan& s, int i = 0)

Example 2D lidar scans (form a venerable SICK LMS200).

Implemented indices: 0,1.

void mrpt::obs::stock_observations::exampleImage(mrpt::img::CImage& im, int i = 0)

Example images (an 800x640 image pair from a Bumblebee 1) Implemented indices: 0,1.

template <class OBSERVATION_T>
OBSERVATION_T::Ptr mrpt::obs::utils::getObservation(
    mrpt::obs::CSensoryFrame::Ptr& observations,
    mrpt::obs::CObservation::Ptr& observation,
    bool priority_to_sf = true
    )

Given an mrpt::obs::CSensoryFrame and a mrpt::obs::CObservation pointer if a OBSERVATION_T type observation is included and return a pointer to that instance.

Pointer to a single instance is going to be returned. If a suitable observation exists in both the CSensoryFrame and CObservation the outcome is decided by the priority_to_sf flag

Returns:

Pointer to the observation of the given type. Otherwise, an empty Ptr object is returned if a valid observation is not found.