[Monory]

There are also more static variants. I know that Yandex collaborated with Zelenograd [0] (city to the north of Moscow) to optimize timings of city traffic lights. Yandex has similar data to what Google does in other places, due to Yandex.Maps app, and they reported to increase average car speed at many intersections without dropping average at any other one.

On the example shown, average car speed at one particular intersection improved from 10 km/h in rush hours to 30 km/h, which is fairly significant.

For many cities which do not want to spend money to replace all traffic lights, such collaborations could be less-than-optimal, but still a great improvement.

[0]: https://yandex.ru/blog/company/38835 (Russian)

[WalterBright]

Using statistics is one way, but is suboptimal. An optimal system would use the cameras to adjust to each situation like a traffic cop would. Though I bet even a modest adaptive light system could do better than a cop, for the simple reason that lights could coordinate with their neighbors.

[narrowrail]

I think your original comment was spot-on, and far more interesting than the hackneyed discussions taking place in this thread. For a little context:

"In 1922 traffic towers were beginning to be controlled by automatic timers. The first company to add timers in traffic lights was Crouse Hinds. They built railroad signals and were the first company to place timers in traffic lights in Houston, which was their home city. The main advantage for the use of the timer was that it saved cities money by replacing traffic officers. The city of New York was able to reassign all but 500 of its 6,000 officers working on the traffic squad; this saved the city $12,500,000."

Crouse-Hinds was still getting sued in the '90s for liability in accidents at intersections. I don't know if these lawsuits are still happening, but I assume liability has something to do with why your suggestion of a more adaptive light system is not tried.