Some features look good in a presentation but become questionable in real traffic.
One of them is showing a pedestrian crossing sign on the HUD.
At first, the idea seems logical. If the system can recognize road signs, why not show the driver everything? Speed limits, no-parking signs, temporary signs, pedestrian crossings. More information, more safety.
But in safety-critical interfaces, that rule is not so straightforward.
Modern Traffic Sign Recognition (TSR) systems in ADAS usually detect signs at a distance of around 50 meters, while more advanced solutions can work at distances of up to 150 meters. This is useful when a driver is moving along a road and has forgotten the current speed limit, missed a no-parking sign, or when the system can show a no-overtaking sign after the driver turns on the indicator to overtake. In these cases, the system reduces cognitive load and helps the driver avoid missing something important.
A pedestrian crossing sign is different.
Here, we need to ask an uncomfortable question: what happens if the driver gets used to treating the system as the source of truth, and the system makes a mistake?
The cost of that mistake is not a fine.
The cost is a person on the road.
The cost is a person on the road.
The Head-up display (HUD) sits inside the driver’s field of attention. Every new element on it competes with the real road scene. Especially if it appears dynamically, is highlighted, animated, or requires immediate reading.
The driver is already processing many things at once: the road, pedestrian movement, nearby cars, traffic lights, signs, mirrors, lane markings, and their own speed.
In the image, I showed a simplified example.
In the first scenario, the driver has four important objects to switch attention between. If each attention shift takes around 0.5 seconds on average, the total reading time is about 1.5 seconds. At 60 km/h, that is around 24.9 meters.
In the second scenario, we add one more icon to the HUD. Now there are five objects. Reading time increases to around 2 seconds. At the same speed, that is already about 33.2 meters.
But that is only the first stage of the problem.
After the driver has read the scene, the mind still needs time to react: recognize the danger, make a decision, and start acting. Even if we assume just 1 second, at 60 km/h the car travels another 16.7 meters during that time.
Then braking begins. In an urban environment, with imperfect road surface, weather, tires, and road conditions, it is reasonable to allow about 35 meters of braking distance before the car comes to a full stop.
So the calculation looks like this:
Scenario without the extra HUD icon:
24.9 m for reading the scene
- 16.7 m for reaction
- 35 m for braking = approximately 76.6 m to a full stop
Scenario with the additional HUD icon:
33.2 m for reading the scene
- 16.7 m for reaction
- 35 m for braking = approximately 84.9 m to a full stop
The difference is about 8 meters.
On the screen, it looks like a small icon.
On the road, it can be the length of several human steps.
And this is not even the most difficult scenario.
Even at 25 km/h, the car is still moving at almost 7 meters per second. If the driver first evaluates one important object, for example a pedestrian on the crossing, and then needs to switch attention to another pedestrian outside the crossing, that attention shift alone can take around 0.5–1 second.
During that time, the car travels roughly 3.5–7 meters before the driver can fully react to the second object.
That is why I did not consider the pedestrian crossing sign a good candidate for HUD display.
The question was not whether we could show it.
We could.
The question was whether we should.
In safety-critical interfaces, not every piece of available information should become visible. Sometimes the best design decision is not to add another signal, but to remove anything that prevents the driver from seeing what matters most.
The best way to help a pedestrian on a crossing is not to show the driver a nice icon.
The best way is not to distract them from the road.
Author
Alexey Dmitriev