Traffic Trainer

Inspiration

We were coming home from the store one night, trying to brainstorm some projects we could do for the public good, when we waited at a light for easily 2 minutes, with no vehicles coming from any other direction. Which for that particular light, isn't terribly uncommon. The idea for a better traffic system came like a bolt of lightning, but since data is gold, we knew we needed a reliable way to generate good data to train on.

What it does

It runs through 86400 seconds (1 day) and (when paired with the right tools) generates random data according to the time of day, and allows a training AI to send signals to the traffic light and pedestrian lights to control traffic. From its actions, cars and pedestrians are then scored on safety and delay, and play into how well it does.

How we built it

We built it on C, a bit of Python, and barely any sleep.

Challenges we ran into

Oo, where to start.. Underestimated probably half of our goals. Everything from python binding to bimodal distribution for car spawning to correlate with time of day. But, we worked through each challenge and didn't give up!