Hacker News Reader

Comment by zelon88 | Hacker News Reader

I'd be happy just to see traffic lights that actually factor traffic congestion. One of my peeves is when a traffic light makes me and a whole line of cars waste brakes to stop then waste gas to sit at a red light just to let one car go when the efficient option would have been to let the line of cars travelling 50mph go first, THEN let the loner who's already stopped go afterwards. It's a problem I've wanted to solve for a loooong time. Imagine the savings in time, gasoline, and environment if every intersection actively tried to maximize efficiency.

View on HN

Not to mention the amount of CO2 emissions that would be reduced from drivers having to stop and re-accelerate fewer times. While I don't know if it would make much of a change during heavy traffic, I think it would make a huge difference during medium to light traffic.

I like the idea of opening up light timings for public review, however figuring out how to optimize everything would probably take a great deal of computational power, and likely out of the scope of local governments that actually make those decisions. It's also not profitable, so I don't see too many companies going after this problem. If only there was a company that had detailed maps, good traffic statistics, insane computational power, and ambitions to do stuff that may or may not be profitable. Maybe they create tools for municipalities to help time their lights.

I agree with most of your points. However, I think it would be easier to use machine learning and computer vision to solve this problem. The profit would come in the form of reduced traffic which would mean more throughput to local businesses. FedEx would profit, trucking would profit, mom-and-pop shops with prime retail locations would profit. Hence, the municipalities would profit from surge in tax revenue. As far as processing goes, the red-light cameras get recorded somewhere, right? Why not double up those servers, put an RPI with a camera on each traffic light, and then beam the data back to the server room for processing. With machine learning and a lot of patience the system would be able to improve itself over time.

I have always believed that this problem could be solved with about $200 worth of hardware at the traffic light, a cheap mobile plan, and a central server location (wherever they host the red-light footage from cameras that already exist).

Now, if anyone has a million bucks laying around I can get to work right away. ;)

San Diego has started to do this. I'm surprised it's taken so long for this kind of technology to get installed given that there's not really any revolutionary technology here.

http://www.sandiegouniontribune.com/news/politics/sd-me-smar...

Reply on news.ycombinator.com