Boundary cues must appear before a collision
Aliases: chaperone · safety grid · play-area warning
What it is
A translucent grid rises from the floor while the toes are still half a step from the wall. A guardian or chaperone boundary has to enter perception before collision, with enough stopping distance left. Arrive too late and the person is still taking the next step on virtual inertia; the grid becomes a post-mortem.
It answers timing and stoppability, not how large the play area is.
Why it happens
The headset hides the real wall. The only cue available for braking is the layer the system draws. From cue onset to feet stopped there is perceptual delay, a judgment that “this is a wall,” and the arrest of an ongoing gait — faster walk, longer stride, longer distance. If the grid fades in only once toes are already past the play area, stopping distance is negative.
The cue also has to beat the virtual world for attention. Low contrast, onset at the view edge, or a look-alike of some translucent effect in the scene, and it is ignored as content. Passthrough that opens the real room in the last half-step is a second channel: it shows the actual wall, but if it opens only on the collision frame it is equally too late. A working boundary is a warning while stopping is still possible, not a close-up of a hit already in progress.
Studying it
Have people walk toward a virtual target across a controlled play-area edge. Manipulate cue distance, style (grid / passthrough / floor line), and walking speed. Observe whether they hit a padded wall, the distance from cue onset to gait stop, and whether they detour rather than slam the brakes. Subjective “felt safe” does not fill this leaf: some people find the grid annoying and hug the edge, with actual braking margin already gone.
Align cue frames to nearest real-object distance with a high-speed camera or headset pose logs. If distance at onset is already less than one step, the trial is a miss regardless of questionnaire scores.
Where it stops holding
Seated, arm-heavy use produces more collisions from hands and controllers; a foot-braking model is not enough, and the cue has to follow the hands. A bystander walking into the play area is a dynamic obstacle the static grid does not cover. If the user drew the play area against the wall, or larger than the real clear space, the system warns against its polygon — into air, or too late. Running and lunging lengthen stopping distance; a half-metre margin tuned for walking will not hold. In a fully dark room passthrough cannot show the wall either, and the grid is the only channel.
Applying it
- Start the boundary at least one stoppable step away, and raise contrast as the edge nears. Do not flash only after the polygon is breached.
- Use grid and passthrough together: grid first, real wall closer in. Both must precede collision, not coincide with it.
- Tune margin to the content’s maximum gait — running or diving passages need a longer onset than a standing menu.
- How to check: with the wall’s location unknown, have people walk toward a virtual target into a real padded wall. They should stop before contact once the cue appears. Contact, or a stop that still hits, means lengthen onset distance, not draw a prettier grid.