Week 5#
Human perception and reaction time at dusk#
You Cannot Brake for What You Have Not Seen#
Last week’s number#
At 80 km/h: 71.3 m total stopping distance.
33.3 m reaction distance + 38.0 m braking distance.
That number starts counting only once the driver has already noticed the animal.
A different question#
What if the animal only becomes detectable 50 metres ahead?
The stopping distance hasn’t changed — it’s still 71.3 m.
What’s changed is whether 71.3 m of road was ever available at all.
Five stages, one boundary#
- the animal is present
- it becomes visible
- the driver recognises it as a hazard
- the driver responds
- braking begins
Week 4’s “reaction time” bundles stages 3 and 4. Stage 2 is this week’s subject.
What changes detection distance#
- contrast — animal against background, not animal alone
- glare — degrades resolution near the light source, doesn’t just dim it
- roadside vegetation — hides outline, independent of light level
- attention and fatigue — visible in principle isn’t the same as noticed
Dusk isn’t universally worse than night. Not every road is the same road.
Distance margin#
Distance margin = assumed detection distance − modelled stopping distance
- positive — the model leaves distance to spare
- negative — stopping before the hazard isn’t possible under those assumptions
Neither result guarantees what actually happens next.
Four scenarios#
Invented teaching inputs — 1,500 kg vehicle, 6.5 m/s² deceleration, same as week 4.
| Scenario | Speed | Detection | Reaction time | Margin |
|---|---|---|---|---|
| A | 60 km/h | 70 m | 1.5 s | +23.6 m |
| B | 80 km/h | 70 m | 1.5 s | −1.3 m |
| C | 80 km/h | 50 m | 1.5 s | −21.3 m |
| D | 80 km/h | 70 m | 2.0 s | −12.4 m |
The margin, compared#
Above the line: distance to spare. Below it: doesn’t clear the modelled stopping distance.
One variable at a time#
- A → B: speed only (60→80 km/h). Margin flips positive to negative — speed moves both terms.
- B → C: detection distance only (70→50 m). Margin worsens 20 m, one-for-one.
- B → D: reaction time only (1.5→2.0 s). Margin worsens 11.1 m — the same figure week 4’s own extended scenario produced.
Reaction time adds distance linearly. Speed hits both terms at once.
Two scene cards#
Scene 1 — open paddock verge. Straight road, open grazed land either side, low orange light, kangaroo side-on at the verge.
Scene 2 — scrub cutting. Same road, dense roadside scrub, kangaroo mostly hidden behind a low bush, oncoming headlights around the bend.
Discussion, not a reaction-time test. Neither scene is calibrated to a real distance.
What this cannot establish#
A calculation using an assumed detection distance is not a measurement of any real driver’s or any real road’s actual detection distance.
It shows how sensitive the model is to an input week 4 held fixed — nothing stronger than that.
Next week#
A distance margin is only as good as the detection distance fed into it — and so far that number has just been assumed.
Next week asks where a real detection distance might actually come from: a sensor, a warning sign, an automatic braking system.
The braking system cannot respond to an animal that is still classified as “probably a shrub.” Your confidence is not an additional headlight.
Physics starts counting after detection. The kangaroo has already been there for some time.