This one is really a bit complicated. It’s experimental, not really deployed; and it is about using crowd-sourced data from lots of smartphone cameras to derive predictions about when traffic lights are going to change, and thereby advice to drivers as to when to slow down so as to avoid having to brake. Braking and starting to move are the aspects of driving that use most fuel, so reducing the number of times that drivers do it saves fuel and carbon emissions.
And the point of using crowd-sourced data is that it would get rid of the need to put expensive smart infrastructure into the traffic lights and then have that communicate with the drivers’ cars.
It has a cool name, of course – SignalGuru. Marketing first, testing later.
This is mainly theoretical at the moment – as far as I can tell, the MIT scientists have just presented a paper about it. Crowdsourcing would only be possible if there were enough drivers with smartphones equipped with the app, and even the authors admit that it wouldn’t be compelling enough without other benefits (like…sigh…prices at different gas stations…the availability of parking spaces).