CogniPilot

MODELICA LIBRARY · ENGINEERING REFERENCE

package

SLAM.Localization

Calibrated relative RGB-D localization and navigation interfaces

From image pairs to pose observations

This layer joins feature processing and calibrated registration into visual observations. Start with ObserveRGBDRelativeFrame for relative-frame processing and RGBDInertialLocalizationInterface for the persistent localization boundary.

RGBDRegistrationUncertainty propagates point noise around the accepted fit. Its covariance is conditional on the correspondences and reference assumptions; it does not automatically include map uncertainty or shared-IMU correlation. Check validity and rejection outputs before forwarding an observation to a filter.

Optical coordinates, body FLU coordinates and world ENU coordinates are separate frames. Preserve calibration, camera extrinsics, capture time and reference identity when adapting an observation.

Members

NameKindDescription
AdvanceFastRGBDLocalizationfunction
AdvanceRGBDLocalizationfunction
InitializeFastRGBDLocalizationfunction
InitializeRGBDLocalizationfunction
ObserveRGBDRelativeFramefunction
RGBDFastInertialLocalizationInitializemodel
RGBDFastInertialLocalizationStepmodel
RGBDInertialLocalizationInitializemodel
RGBDInertialLocalizationInterfacemodel
RGBDInertialLocalizationStepmodel
RGBDLocalizationCatalogpackage
RGBDLocalizationFramepackage
RGBDLocalizationOrientationmodel
RGBDLocalizationProcessingpackage
RGBDLocalizationTrackingfunction
RGBDOpticalPointCovariancefunction
RGBDProperRotationmodel
RGBDProperRotationValuefunction
RGBDRegistrationSandwichfunction
RGBDRegistrationUncertaintymodel
RGBDRelativeBodyPosefunction
RGBDRelativePosemodel
RGBDUncertaintyInverse6function
RGBDUncertaintyProperfunction
RGBDUncertaintySkewfunction
RGBDVisualObservationmodel
RGBDVisualRelativeObservationmodel

Modelica declarations and implementation

Show parameters, interfaces and source

Parameters, inputs, outputs, defaults and equations from the source file.

package Localization
  "Calibrated relative RGB-D localization and navigation interfaces"
  annotation(Documentation(info = "<html>
    <h4>From image pairs to pose observations</h4>
<p>This layer joins feature processing and calibrated registration into visual
observations. Start with
<a href=\"modelica://SLAM.Localization.ObserveRGBDRelativeFrame\">ObserveRGBDRelativeFrame</a>
for relative-frame processing and
<a href=\"modelica://SLAM.Localization.RGBDInertialLocalizationInterface\">RGBDInertialLocalizationInterface</a>
for the persistent localization boundary.</p>
<p><a href=\"modelica://SLAM.Localization.RGBDRegistrationUncertainty\">RGBDRegistrationUncertainty</a>
propagates point noise around the accepted fit. Its covariance is conditional
on the correspondences and reference assumptions; it does not automatically
include map uncertainty or shared-IMU correlation. Check validity and rejection
outputs before forwarding an observation to a filter.</p>
<p>Optical coordinates, body FLU coordinates and world ENU coordinates are
separate frames. Preserve calibration, camera extrinsics, capture time and
reference identity when adapting an observation.</p>
    </html>"));
end Localization;