Revision history [back]

It either has to exist separately in parallel under a different name like 'ar_tag_camera', or you need a node that can fuse the camera information with the other sensor input to provide a higher quality position than the EKF or camera by themselves (or equivalently the EKF needs to take in the camera position as an input). Maybe there is a way to shoehorn the camera position to look like some other already supported sensor (it is roughly like gps), but the likelihood of getting that to work is probably lower than if the software was designed for a camera looking at a ar tag or similar.