Augmented Windshield
2019 · RIT Information Science and Technology (IST) academic project
Academic study of how far from the road a heads-up widget can sit before the driver stops noticing the car ahead braking.
An academic project asking a question that has aged well: as windshields turn into displays, where can you actually put information without degrading the driving? The apparatus was a VR driving task built in Unity, with a head-mounted eye tracker recording where the driver looked while a lead vehicle braked at unpredictable moments. I have rebuilt that apparatus as something you can run in a browser, so the manipulation is visible rather than described.
Live demoRun the driving task yourself: watch the car ahead, hit the brake key when it lights up, then move the widgets further into the periphery and watch your own reaction time change.Open the live demoInteractive prototype
The problem
A heads-up display has one unavoidable tension. Put information where the driver is already looking and it occludes the road; move it out of the way and it stops being read at all. The project needed a way to measure that trade off rather than argue about it, which meant building a driving task where the position of the information was the only thing that changed.
Process & prototyping
The original apparatus (Unity, VR)
- A lead vehicle that never settlesThe car ahead oscillates in depth continuously, so the driver has to keep attending to it instead of parking their gaze somewhere convenient. The range of that oscillation was a tunable parameter.
- Brake events as the measureAt unpredictable intervals the lead vehicle brakes and its lights go red. Time from that moment to the driver reacting is the number the whole apparatus exists to produce.
- Widgets spawning on the windshieldHeads-up elements appear a few times a second at positions drawn from the condition under test. Realistic in kind, controllable in where they sit.
- Eye tracking alongsideA head-mounted Pupil Labs tracker recorded gaze through each session, so where the driver actually looked could be set against how quickly they reacted.
The browser rebuild
- The same task, minus the hardwareThe oscillating lead vehicle, the brake events, the widget spawner and eccentricity as the control all carry over. The headset and the eye tracker do not, so nothing pretends to replace them.
- Eccentricity is the one controlA single slider moves the widgets from directly over the road out to the pillars, because that is the variable the study was actually about.
- Your numbers, not the study'sReaction times are measured from your own responses, banded by eccentricity, with misses counted separately. Nothing is stored or sent, and no figure is carried over from the original sessions.