Research & Papers

Your Rental E-Scooter May Soon Know If You're Too Drunk to Ride

⚡Sensors caught drunk riders with near-perfect accuracy — without a breathalyzer.

Deep Dive

Researchers in Sweden built a special e-scooter covered in sensors and asked 25 volunteers to ride it around a test track three times: once sober, once at a blood alcohol level of 0.05%, and once at 0.08% — roughly the legal drunk-driving limit in most U.S. states. The sensors measured lean, steering, throttle and braking about 100 times per second, capturing exactly how the scooter moved under each rider.

The finding is simple and a little surprising. Drunk riders didn't wobble randomly — they changed style. Sober riders make constant tiny steering corrections, keeping the scooter smooth and steady. After a few drinks, those small corrections mostly disappear, replaced by fewer but much bigger, jerky overcorrections. It's the same pattern you see in a drunk driver swerving and then yanking the wheel back. Side-to-side sway and steering speed were the clearest giveaways.

A computer program then read those motion patterns and guessed each rider's state. It got the answer right 85% of the time overall, and when comparing a sober ride to a heavily drunk one, it was perfect — 100% accurate. That matters because today's tools are blunt: city curfews, in-app quizzes, or outright bans after certain hours. Those punish everyone equally. A scooter that can sense impairment in real time could let responsible riders keep riding at 2 a.m. while quietly stopping the ones who shouldn't be on the road.

The catch: this was 25 people on a closed test track. No traffic, no rain, no cobblestones, no one swerving to avoid a pedestrian — all things that could confuse the system in the real world, and could also flag sober riders who are just riding badly. There's also a privacy question: do you want your rental scooter recording and reporting how you ride? And any system that locks people out would need clear rules, appeals, and probably government approval before it shows up on a scooter near you.

Key Points
  • An e-scooter loaded with motion sensors could tell drunk riders from sober ones with 85% accuracy — and 100% accuracy at the highest alcohol level tested.
  • Drunk riding looks different, not just worse: riders stop making small steady corrections and start making big jerky ones.
  • The study only involved 25 people on a closed test track, so real streets, weather and traffic could still trip the system up.

Why It Matters

Could make late-night scooter rides safer without punishing sober riders — if cities and scooter companies adopt it.

📬 Get the top 10 AI stories daily