package

SLAM.Fusion

Visual measurements for the CogniPilot navigation estimator

Independent visual corrections for the right-error navigation ESKF.

correctMappedLandmarksLoose compresses RGB-D observations into a six-dimensional pose statistic and requires sufficient geometric rank.
correctMappedLandmarksTight uses pixel/depth residuals directly and can retain partial constraints.

Both paths assume a fixed known map and independent new camera noise. Select visible observations and the matching rows and columns of their full covariance before calling either function; neither consumes a visibility mask. Uncertain live maps and external poses sharing IMU/GPS inputs require additional map uncertainty and cross-correlation treatment.

Use LandmarkCamera for synthetic inputs and VisualNavigationReplay for sequential GPS and camera aiding.

Members

Name	Kind	Description
linearizeMappedLandmarks	function	Shared RGB-D residuals for fixed, known map landmarks
compressPose	function	Sufficient pose statistic at one common camera linearization
correctMappedLandmarksLoose	function	Fuse a compressed pose from independent fixed-map RGB-D observations
correctMappedLandmarksTight	function	Fuse raw independent fixed-map RGB-D observations

Modelica declarations and implementation

Show parameters, interfaces and source

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

package Fusion "Visual measurements for the CogniPilot navigation estimator"
  annotation(Documentation(info = "<html>
    <p>Independent visual corrections for the right-error navigation ESKF.</p>
    <ul>
    <li><a href=\"modelica://SLAM.Fusion.correctMappedLandmarksLoose\">correctMappedLandmarksLoose</a>
    compresses RGB-D observations into a six-dimensional pose statistic and
    requires sufficient geometric rank.</li>
    <li><a href=\"modelica://SLAM.Fusion.correctMappedLandmarksTight\">correctMappedLandmarksTight</a>
    uses pixel/depth residuals directly and can retain partial constraints.</li>
    </ul>
    <p>Both paths assume a fixed known map and independent new camera noise.
    Select visible observations and the matching rows and columns of their
    full covariance before calling either function; neither consumes a
    visibility mask. Uncertain live maps and external poses sharing IMU/GPS
    inputs require additional map uncertainty and cross-correlation treatment.</p>
    <p>Use <a href=\"modelica://SLAM.Simulation.LandmarkCamera\">LandmarkCamera</a>
    for synthetic inputs and
    <a href=\"modelica://Tests.VisualNavigationReplay\">VisualNavigationReplay</a>
    for sequential GPS and camera aiding.</p>
    </html>"));
end Fusion;