3 ms·
Seems there is an important opportunity here to apply machine learning to video feeds of incoming traffic and people flow, and shape the phases to minimize wait
by yholio 8y ago
Seems there is an important opportunity here to apply machine learning to video feeds of incoming traffic and people flow, and shape the phases to minimize wait time for all citizens, equally. The current systems require expensive induction loops and only care about cars and button pushers, this being how we got here.
If people learn the system acknowledges their needs and tries it best to let them cross safely, it might lead to much better long term compliance then some fake buttons which are bound to be exposed as fake. The whole idea of giving higher priority to cars is an insult to taxpayers walking their neighborhoods.
- shereadsthenews 8y agoIt would be better to just build smaller intersections that do not need millions of dollars worth of signals and cameras and ML inference rigs.
- yholio 8y agoThe cost of smart trafic lights is peanuts compared to the costs of reconfiguring and rebuilding a major intersection.
- noobiemcfoob 8y agoYour comment is based on no data at all and just a handful of assumptions that are most certainly far from true.
- yholio 8y agoI happen to work for a council and know the general order of magnitude for road infrastructure investments - I participate in their approval. I would file this suggestion of "building smaller intersections", that is, completely re-routing the traffic flows in a given area, in the purely fantastical bin. If we could squeeze 20% more traffic capacity from a given congested intersection with a one time investment of, say, under €200K, it would be a major breakthrough. We use a complex system that involves periodically measuring traffic flows using pneumatic tubes, then feeding that data to a computerized model that will produce the timing sequences for the lights lights and their evolution during a day. There is a crew moving around the district installing these tubes and counters for a week or so. The current state of the art are induction loops embedded in the asphalt that measure the number of incoming vehicles and adjust the timings in real time. A purely optical detection that can work in all weather and recognize all types of vehicles or even pedestrians would be quite a feat. Anyway, thank you for your factual and data infused contribution.
- noobiemcfoob 8y agoHahaha well that's all you had to say! You're more informed than your first comment led me to believe. Thanks for writing this out. If you wish to avoid my disparaging remarks in the future, lead with this and not vacuous negativity on someone else's thoughts.