If your questions aren't answered here, please feel free to make an issue!
FAQ
Localization
What kind of accuracy can I expect?
Depends on the "quality" of the created anchor and how close the viewpoints that were used to create the anchor are to the viewpoints querying. In the best case, expect 1-5 cm and about a degree error. In the "average" case, expect ~10+ cm and a few degrees of error.
This is meant to be sufficient for human-robot interaction with Mixed Reality, but not for accuracy-critical applications.
Advice for how to create anchors?
Try to get as much information about the surroundings in the session as possible before calling create_anchor. This means many different viewpoints, attempting to get full coverage of the space, etc.
If you're creating anchors on, for example, a mobile device and querying from a robot, make sure to get representative viewpoints for the robot.
Make sure you're moving slowly enough that the tracking of your SLAM system doesn't diverge.
Known Issues
Will be added here once we have some known issues. :)