World-locked content stays fixed in the environment

What it is

A virtual nameplate on the side of a machine stays on that door as you walk a full circle around it. That is world-locked content: its pose is written in the environment frame, not on the torso and not on the head. Walk away and it leaves the view; look back and it is still there. The question it answers is which point in this room the information belongs to, not which patch of the retina it should occupy right now.

Why it happens

Six-degree-of-freedom tracking places the headset under a world origin. World-lock parents the interface transform to that origin: each frame the projection is contentWorldPose × inverse(hmdPose). Translation produces motion parallax — near panels shear against the background, far ones barely move — the same signature real objects have. A turn sends the panel out of view along the room geometry, the way a fire extinguisher on the wall would leave.

Stationarity is an achievement of tracking, not of the pixels. When tracking drifts, the nameplate slides along the real wall; when tracking drops, it jumps with the wrong origin. Because the environment is the reference, a moved chair or a changed light leaves the content sitting on yesterday’s coordinates, looking as if it floats.

Studying it

Hang the same panel in world coordinates and in headset coordinates, then run a spatial-memory task: walk around the object and later point to where the information was. World-lock should still be pointed back to the original place; the other parenting takes the answer with the person.

Independent variables: reference frame of the content, walking radius, whether a turn happens in between. Dependent variables: pointing error, time to re-find the panel, the spatial-presence subscale of IPQ.

An empty lab overestimates how “object-like” world-lock feels, because there is nothing to occlude it. In a furnished room, people remember the panel by its relation to a table edge.

Where it stops holding

World-lock only reads as “nailed to the room” while positional tracking is stable. A 3DoF headset has no translation, so world-lock collapses into a painting aimed at the initial heading and misregisters the moment someone stands up. On optical see-through, a geometric mismatch between a virtual nameplate and a real wall is more jarring than a number glued in view, because the nameplate claims to be part of the room. For seated, unturning viewing, the three lock modes look similar for a short time; world-lock’s advantage shows up after walking.

Applying it

• Put information that belongs to the environment — machine labels, a room schedule, files that live on a desk — into world coordinates. Do not let it follow a turn of the visor.
• Acceptance: walk at least a two-metre radius around the content. The panel must keep its relation to walls and tables, and must not yaw with the headset.
• When tracking is lost, do not pretend the nail still holds — freeze the last world pose and mark it unreliable, rather than sliding in silence.
• How to check: place a world-locked panel in a real room, blindfold someone who did not lay it out, spin them twice, and ask them to point. Pointing into empty furniture gaps, or pointing in front of their own chest, means the reference frame was not the environment.

Related

• Same group: N3.01.2 Body-locked content moves with the person but not with head turn · N3.01.3 Head-locked content stays in the same place in the view · N3.01.4 Head-lock is the most uncomfortable mode for prolonged use
• Nearby: N3.08 World-locked, Body-locked, and Head-locked · N2.07 The Body as Reference Frame
• Search terms: world-locked · world-referenced UI · spatial registration