Google has shared an update about its AI-driven mission to reduce wait times, fuel use, and polluting emissions at intersections – and the early indications are that stops can be reduced by around 30 percent, with emissions dropping around 10 percent.
Just network the sensors. Let intersections know about adjacent intersections. Anything more distant than that doesn’t exist.
You don’t need any intense city-wide apparatus, or a big-iron traffic brain flipping every light from afar. (Though internet access might be the cheapest way to move the data to and from each intersection.)
Just… stop letting a red light… trigger another red light, further down. That’s what happens when a crossroad waits for a lull in a main road. Usually that lull is from a red light, nearby. And if only a few cars cross, each red will be pretty short, and the lull will move along with the cars, and those drivers will wish everything evil in the world happens to you, and your little dog too.
If an adjacent intersection just turned green, and this intersection is green for them, keep it that way, at least until that bulk of cars tapers off.
Other than that - yeah, roundabouts. Controlled intersections should be the last resort, and we have the technology to let sparse traffic just blow through. It’s a circle.