BikeBot

If you've ever biked on Stanford's campus after dark, you know the drill: dodging pedestrians, hoping you didn't forget your bike light, praying your bike does not get randomly stolen in the middle of the night.

Here's the thing: Stanford takes bike safety seriously. It's practically a Stanford tradition at this point that you turn a corner and see campus cops ensuring your lights are on in the dark.

On the other hand, campuses are a hot bed for bike thefts, and that sucks especially when you are a broke undergrad.

So we built the robot that does it better, making sure students stay safe, and their bikes safe as well!

Meet BikeBot

BikeBot is an autonomous patrol robot that roams campus, detects cyclists in real-time using AI vision, and flags anyone riding without proper lighting while tracks light Here's how it works:

Running on a Jetson Orin Nano, BikeBot uses NanoOWL, an on-device open-vocabulary detection model, to spot bikes from a live camera feed.

🔦 Detects whether a bike light is present (or suspiciously absent) 📹 Clips a 5-second recording as evidence 📄 Generates a full incident report automatically 💬 Lets safety staff query incidents via an AI chat agent on the live dashboard 🔦 Tracks bikes history

The robot itself is driven by two brushless motors controlled through an Arduino

Why This Matters Every year, dozens of cycling accidents on campuses happen at night due to poor visibility. Bike lights aren't just a rule, they're the difference between a close call and an ER visit. Current patrol methods are manual, and sparse. BikeBot makes enforcement data-driven.

What's Next Autonomous navigation to high-incident hotspot, so the robot knows to stake out the Oval at 11pm on a Friday, because that's just statistically where the lightless cyclists are.

Built With

• ai
• arduino
• cv
• depthanything-model
• imu
• jetson-nano
• llama
• nvidia
• openstreetmap
• vllam