Several people need a shared coordinate reference

What it is

Two people both point at “the cube on the table,” and the cube sits on two different tables. Each headset treats its own boot pose as the origin; without a prior alignment, the room is not the same room. A shared coordinate system is the premise of co-located work: several people have to read “there” against one origin and one set of axes before a virtual object can land on one physical surface.

Without a common reference, collaboration is not a bit worse. The referent does not even live in one world.

Why it happens

Each device’s session origin is local: stance, heading, and floor estimate can all differ. Objects placed against those origins fork in physical space. Alignment turns one device’s pose — or one shared anchor — into everyone else’s reference: scan a code, look together at one featured surface, or stand on a floor mark for a calibration. Alignment error grows with distance — a centimetre on the desktop becomes a handspan on the far wall. If each headset estimates gravity on its own, floor height splits too, and “your floor” lands at someone else’s shin.

A shared frame answers whether “there” is the same there. It does not answer who is facing where, or whose state is newer on the network. Those questions start after alignment.

Studying it

A co-located placement task: each of two people seats the same object “at the centre of the real table” in their own headset. Measure the physical distance and yaw between the two objects with an external tape. Alignment method (code, shared view of a surface, floor footprints) is the condition. Success is physical coincidence, not a self-report of “we see the same thing.”

Independent variables: alignment method, standing distance between the two people, whether floor estimates are independent. Dependent variables: translation gap, yaw gap, later hit rate when pointing at “the same object.”

One person watching a 2D mirrored view does not test whether coordinate systems aligned — the mirror already projected the coordinates away.

Where it stops holding

Remote collaboration has no shared physical room; a common frame is either absent or belongs to some other virtual room, and co-located calibration does not transfer. One person demoing while a neighbour watches a monitor also needs no two-headset alignment. A single user has one origin by construction. Once alignment is done, each tracker can hold it for a while; over time they drift apart and the common reference undoes itself — that is a later layer, not the “must we align” layer.

Applying it

• Put a visible alignment step before anyone starts operating. Do not assume boot is enough to share a table.
• Accept with a real prop: both people set a virtual cross on this coaster and check it is one physical point.
• On alignment failure, stop at the collaboration door. Do not drop people into a room where each is speaking of a different there.
• How to check: each person indicates where “the cube’s corner” sits on the real table; measure the two points. Past what the task can tolerate, there is still no shared coordinate.

Related

• Same group：N5.05.2 Each has a different viewpoint, so reference must be made explicit · N5.05.3 Permissions on shared content must be clear
• Nearby：N5.10 Shared Space and Multi-user Collaboration · N3.08 World-locked, Body-locked, and Head-locked
• Search terms：shared coordinate system · colocation · spatial alignment