An overlay hiding a dynamic hazard in the real environment is the highest-grade risk

What it is

A virtual exploded view of a part hides a forklift that is reversing. A hidden pillar is bad; the pillar will not appear under your feet. Something that moves itself and can change heading, hidden, is dynamic hazard occlusion — the top grade on a safety ladder. People, vehicles, robot arms, a fork tine, a running child all sit on that grade.

Static occlusion is missing a map. Dynamic occlusion is not having time to yield.

Why it happens

Avoidance depends on seeing an approaching contour and heading inside a time-to-collision window. Once an overlay covers that patch of retina, the window has no update. A static object can be patched from memory (“a pillar there”); a dynamic object’s position changes inside the window, and memory patches a stale coordinate. Hazards also often enter from the edge of the view, while overlays like the centre and high contrast — stealing the first fixation that belonged to the oncoming vehicle, or fracturing the edge contour.

The grade comes from irreversible consequence: a missed torque can be redone, a missed forklift cannot. So this does not negotiate with “a little denser, a little thinner.” If an overlay and a dynamic hazard intersect in the line of sight, that is a fault, not a style.

Studying it

Controlled approachers: during walking or manipulation, a real, safely contactable moving object (a person, a slow vehicle mock, an extending pole) enters the path through the overlay region. Compare no overlay, translucent overlay, solid overlay, and a “cut out toward the hazard” condition. Primary scores: detection distance, time to yield, collisions/contacts.

Motion factors: speed, entry direction (ahead / side), whether the layer fully covers. Overlay factors: opacity, area, depth cut-out or not. Detection: verbal “I saw it,” first fixation, moment the body starts to yield.

An animated person on a screen in place of a real approacher drops collision fear and body response; use it only as a pre-test.

Where it stops holding

A fully static bench with no third person and no moving machinery may not trigger the top grade for a while; still treat “someone walks in” as an event that can fire at any time, not as a permanently sealed cell. Video see-through that can hide the real behind virtual geometry is harsher here than optical see-through’s leaky additive light — the stronger the capability, the harder this rule. Distant, slow approachers that hearing can cover (a cart in a corridor) lower visual dependence; that is not a forklift policy on a shop floor. Children and pets move less regularly; a constant-speed dummy in the lab will understate.

Applying it

• Classify people, vehicles, and robot arms as never-cover: no solid layer in their direction, or cut the layer out by depth.
• When dynamic objects cannot be reliably detected, assume they may appear from any direction: cap solid area in the centre, keep the lower and side view.
• On detecting an approacher, tear down intersecting overlays immediately. Do not fade — the fade is still playing when the vehicle arrives.
• How to check: while the wearer looks at a solid part drawing, have a person enter the path from ahead-left. They should be seen and yielded to outside contact distance. Discovery at a brush or a collision means the drawing is sitting on the top-grade risk.

Related

• Same group：N5.13.2 While moving, overlay density should drop to leave attention for the environment · N5.13.3 On a performance crash or see-through failure, fallback must prioritize environment visibility · N5.13.4 Stairs, crossings and other safety-critical zones need a rule that clears or fades overlays · N5.13.5 Long-term reliance on overlay cues weakens active scanning of the real environment
• Nearby：N5.07 Overlay Density · N5.12 Overlay Density in Real-world Context
• Search terms：dynamic hazard occlusion · moving obstacle · time-to-collision