Detailed Description

Version:: 2.0

Author:: thomas.lemaire@laas.fr

SLAM (Simultaneous Localization And Mapping) is a fundamental problem for robotics.

The module slam implements several algorithms:

The classic SLAM algorithm based on a extended kalman filter
An original algorithm for the special bearing-only SLAM problem, a multi gaussian distribution is used to approximate the initial state. This is a delayed initialization algorithm.
The inverse-depth monocular SLAM algorithms. This is an undelayed initialization algorithm
A delayed monocular SLAM with segments.
An undelayed monocular SLAM with segments.

Currently, sizeless points, iinverse-depth points and Plucker lines can be used as features.

History

0.7 (2007-03-07) - port to datareader
0.6 (2006-07-10) -
0.5 (2006-05-29) - smart selection of robot poses in boSLAM, points at inifinity are now considered, implementation of "inverse-depth" parametrization for boSLAM
0.4 (2006-04-18) - boSLAM using an omni camera
0.3 (2005-12-15) - boSLAM with segments
0.2 (2004-12-15) - Bearing-only SLAM
0.1 (2004-10-18) - Initial version, basic slam based on an EKF, point features

Overview

Main classes

jafar::slam::SlamEkf implements the classic EKF based SLAM algorithm
jafar::slam::BearingOnlySlam implements the bearings-only SLAM algorithm described in my thesis (Thomas Lemaire thesis)
jafar::slam::BaseFeature is the feature container
jafar::slam::FeatureModel defines how to compute a landmark state vector in a different frame
jafar::slam::FeatureObserveModel defines an observation model for a given sensor and type of landmark
jafar::slam::BearingOnlyFeatureObserveModel defines an observation model for a given bearings sensor and type of landmark

Example of use (in tcl)


#
# setup
#

# create a Slam object
set slamEkf [slam::new_SlamEkf 350 6 6]

# optionnaly change the map manager
#   $slamEkf setMapManager [slam::new_LocalMapManager $slamEkf 40 3]
#   $slamEkf setMapManager [slam::new_GlobalMapManager $slamEkf 3 3.0]

# covariance update method (DEFAULT, STANDARD, JOSEPH)
[$slamEkf getFilter] setCovUpdateType $filter::BaseKalmanFilter_STANDARD
# consistency check method
[$slamEkf getFilter] setupConsistencyCheck $::filter::BaseKalmanFilter_CONSISTENCY_EXCEPTION 10.0


# create a landmark model
set ptModel [slam::new_PointFeatureModel]
# create an associated observe model
set ptObsModel [slam::new_PolarPointFeatureObserveModel $ptModel]
$ptObsModel setNoiseValues 0.1 [jmath::degToRad 2] [jmath::degToRad 2]
# add the defined sensor to Slam
$slamEkf setSensor $ptObsModel
# set the robot to sensor t3d
set robotToSensor [jmath::new_vec 6]
jmath::setValue $robotToSensor "(0,0,1.0,[jmath::degToRad 90],0,0)"
$slamEkf setRobotToSensor $robotToSensor

# create a prediction model
set odoPredictModel [slam::new_Odo3dPredictModel]
[$odoPredictModel cget -odoNoiseModel] set [expr pow(0.05,2)] 0 [expr pow(0.05,2)] 0

#
# run
#

# predict with command u [v,w]
$slamEkf predict $predictModel $u

# update
FIXME

Example of use (in tcl)


# create and initialize the slam object
slamEkf = Slam::SlamEkf.new(350, 6, 6)
slamEkf.getFilter.setCovUpdateType(Filter::BaseKalmanFilter::STANDARD)
slamEkf.getFilter.setupConsistencyCheck(Filter::BaseKalmanFilter::CONSISTENCY_EXCEPTION, 10.0)

# create a landmark model
ptModel = Slam::PointFeatureModel.new
# create an associated observe model
ptObsModel = Slam::PolarPointFeatureObserveModel.new( ptModel )
ptObsModel.setNoiseValues( 0.1, Jmath::degToRad(2), Jmath::degToRad(2))
# add the defined sensor to Slam
slamEkf.setSensor(ptObsModel)
# set the robot to sensor t3d
robotToSensor = Jmath::Vec.new(6)
Jmath::setValue(robotToSensor, "(0,0,1.0,#{Jmath::degToRad(90)},0,0)")
slamEkf.setRobotToSensor(robotToSensor)

# create a prediction model
odoPredictModel = Slam::Odo3dPredictModel.new
odoPredictModel.odoNoiseModel.set(Math::exp( 0.05 ** 2 ), 0, Math::exp( 0.05 ** 2 ), 0 )

#
# run
#

# predict with command u [v,w]
u = Jmath::Vec.new(2)
slamEkf.predict(odoPredictModel, u)

# update
#       FIXME

Macro

Extra doc for macro can go here... (you can delete this section if not relevant)

Tcl interface (generated by swig)

The interface of the module is generated from the following files:

slam.i defines the wrapped classes and functions,
slamException.i defines the try { } catch block for this module.

Classes
class	jafar::slam::AbstractMapManager
	Abstract map manager. More...
class	jafar::slam::AbstractMapObject
	class Abstract Map Object More...
class	jafar::slam::BaseSlam
	Base class for Slam algorithms that can be used by managers. More...
class	jafar::slam::BasisFeatureModel
	A basis features includes the X,Y,Z coordinates of the origin, plus the raw, yaw and pitch angles. More...
class	jafar::slam::InitStateMember
	Initial state gaussian member (or hypothesis). More...
class	jafar::slam::InitFeature
	A generic bearing only feature. More...
class	jafar::slam::BearingOnlyFeatureObserveModel
	Bearing only feature observation model. More...
class	jafar::slam::InfFeatureObserveModel
	A feature at infinity observe model. More...
class	jafar::slam::TrajectoryPoseRecord
	A pose from the trajectory. More...
class	jafar::slam::BearingOnlySlam
	Bearing only slam. More...
class	jafar::slam::DalaManager
	Class for managing Dala. More...
class	jafar::slam::BaseFeature
	Generic feature. More...
class	jafar::slam::FeatureModel
	This class defines a generic feature model. More...
class	jafar::slam::BaseFeatureObserveModel
	The root class for all feature models. More...
class	jafar::slam::FeatureObserveModel
	Fully observable feature observe model. More...
class	jafar::slam::Full3dPredictModel
	Full 3d predict model. More...
class	jafar::slam::FreeStateCollector
	This class acts as a collector of free state in the SLAM. More...
class	jafar::slam::DefaultMapManager
	Default map manager. More...
class	jafar::slam::LocalMapManager
	Local slam manager. More...
class	jafar::slam::GlobalMapManager
	Global (long term) map manager. More...
class	jafar::slam::Observation
	Observation. More...
class	jafar::slam::OdoNoiseModel
	Odometry noise model. More...
class	jafar::slam::Odo3dPredictModel
	3d odometry predict model. More...
class	jafar::slam::PointFeatureModel
	Sizeless 3D point model. More...
class	jafar::slam::CartesianPointFeatureObserveModel
	sizeless point cartesian observe model. More...
class	jafar::slam::PolarPointFeatureObserveModel
	sizeless point polar observe model. More...
class	jafar::slam::BearingPointFeatureObserveModel
	sizeless point bearing-only observe model. More...
class	jafar::slam::ImagePointFeatureObserveModel
	sizeless point from perspective image observe model. More...
class	jafar::slam::StereoImagePointFeatureObserveModel
	sizeless point from stereo images observe model. More...
class	jafar::slam::OmniImagePointFeatureObserveModel
	sizeless point from omnidirectional image observe model. More...
class	jafar::slam::PointInfFeatureModel
	Sizeless 3D point at infinity model. More...
class	jafar::slam::BearingPointInfFeatureObserveModel
	Point at infinity bearing-only observe model. More...
class	jafar::slam::OmniImagePointInfFeatureObserveModel
	Point at infinity observed in a panoramique image. More...
class	jafar::slam::PointInvDepthFeatureModel
	Sizeless 3D point using the inverse-depth parametrisation. More...
class	jafar::slam::BearingPointInvDepthFeatureObserveModel
	Sizeless 3D point using the inverse-depth parametrisation, bearings-only sensor. More...
class	jafar::slam::OmniImagePointInvDepthFeatureObserveModel
	Sizeless 3D point using the inverse-depth parametrisation observed in a panoramic image. More...
class	jafar::slam::ImagePointInvDepthFeatureObserveModel
	Sizeless 3D point using the inverse-depth parametrisation observed in a perspective image. More...
class	jafar::slam::PoseCopy
	A special object of the map to estimate a local frame. More...
class	jafar::slam::BaseRobot
	Base robot model. More...
class	jafar::slam::SegmentAnchorFeature
	A plucker segment anchored feature, with its extremities. More...
class	jafar::slam::SegAnchorPluckerFeatureModel
	3D segments using anchored plucker parametrisation. More...
class	jafar::slam::ImageAnchorPluckerFeatureObserveModel
	3D segments using the anchored plucker parametrisation observed in a perspective image. More...
class	jafar::slam::SegmentObservation
	Observation of a segment. More...
class	jafar::slam::StereoSegmentObservation
	Observation of a segment. More...
class	jafar::slam::SegmentFeatureContraintModel
	Segment Feature constriant model. More...
class	jafar::slam::SegmentFeature
	A segment feature, with its extremities. More...
class	jafar::slam::SegmentFeatureModel
	segment model. More...
class	jafar::slam::ImageEuclideanPluckerFeatureObserveModel
	segment extracted from an image observe model. More...
class	jafar::slam::ImagePluckerFeatureObserveModel
	Sizeless 3D segment using the homogenous coordinates parametrisation observed in a perspective image. More...
class	jafar::slam::RhoThetaImagePluckerFeatureObserveModel
	Observation Model using Rho Theta parametrisation. More...
class	jafar::slam::ScaledHomogeneousImagePluckerFeatureObserveModel
	Observation Model using Scaled Homogeneous parametrisation. More...
class	jafar::slam::StereoImagePluckerFeatureObserveModel
	Observation Model using a Stereo head, two uvd points for init, two rho-thetas for updates. More...
class	jafar::slam::SegmentIDFeature
	A segment inverse depth feature, with its extremities. More...
class	jafar::slam::SegInvDepthFeatureModel
	3D segments using 2 points inverse-depth parametrisation. More...
class	jafar::slam::ImageSegInvDepthFeatureObserveModel
	Sizeless 3D point using the inverse-depth parametrisation observed in a perspective image. More...
class	jafar::slam::ImageSegInvDepthInovFeatureObserveModel
	Sizeless 3D point using the inverse-depth parametrisation observed in a perspective image. More...
class	jafar::slam::SlamEkf
	This class solves SLAM problem using the EKF-based academic solution. More...
class	jafar::slam::SlamEventListener
	Interface for Slam listeners. More...
class	jafar::slam::SlamEventAdapter
	Adapter for Slam listeners. More...
class	jafar::slam::BoSlamEventListener
	Interface for BoSlam listeners. More...
class	jafar::slam::BoSlamEventAdapter
	Adapter for BoSlam listeners. More...
class	jafar::slam::SlamException
	Base class for all exceptions defined in the module slam. More...
class	jafar::slamlines::ImageSegment
	class Image Segment More...
class	jafar::slamlines::ImageSegmentManager
	class Image Segment Manager More...
class	jafar::slamlines::ImageHomogenCoordSegmentManager
Namespaces
namespace	jafar::slam::EulerTools
	Functions for operations related to Euler angles lines in SLAM.
namespace	jafar::slam::lineTools
	Functions for operations related to lines in SLAM (Plucker, homogeneous, rho-theta, and others)
Files
file	eulerTools.hpp
	Toolbox for operations related to Euler angles in SLAM.
file	lineTools.hpp
	Toolbox for operations related to lines in SLAM (Plucker, homogeneous, rho-theta, and others)
Functions
bool	jafar::slam::operator< (TrajectoryPoseRecord const &p1, TrajectoryPoseRecord const &p2)
	This operator compares TrajectoryPoseRecord.

Function Documentation

bool jafar::slam::operator<	(	TrajectoryPoseRecord const &	p1,
		TrajectoryPoseRecord const &	p2
	)

This operator compares TrajectoryPoseRecord.

It is used to find the pose to throw away.

Detailed Description

History

Overview

Main classes

Example of use (in tcl)

Example of use (in tcl)

Macro

Tcl interface (generated by swig)

Classes

Namespaces

Files

Functions

Function Documentation