If each traffic light responds to its immediate demands, then all the lights will just react to the traffic coming from nearby intersections, which defeats the purpose of a smart network.
The solution is a decentralized approach that lets the traffic lights work together by figuring out how changes at each individual intersection would affect the entire system. Instead of being stymied by natural fluctuations in traffic, the system takes advantage of them, using random gaps to help improve traffic flow. Traffic lights could request green time only when there is a definite demand for them, the researchers write. This acyclic approach could eliminate the particularly annoying problem of sitting at a red light while there’s no traffic.
http://www.popsci.com/cars/article/2010-09/giving-traffic-lights-mind-their-own-can-reduce-congestion-study-says | Previously: Is a Progress Bar on a Traffic Light a Good Idea? | Photo by Flickr user grendelkhan used under Creative Commons license
Ech traffic position is monitored and the data fed back to the central computer(s).
If a traffic light loses communication with the central computers, then it takes over its own control knowing the day of the week, the time and draws upon its "collected" knowledge of what needs to be done and how....
But it is also checking the comms and opnce communication is achieved with the central computer(s), it reverts back.....
To get traffic moving properly through a large city, the central control is needed....
I think I can speak for many of those in the US when I say 'bad idea'. We typically drive vehicles larger then bicycles. Last thing I want to do on a trip through town is to meet somebody in my mid-size sedan (read 'large' if in Europe) and them in their Suburban (read 'bus' if in Europe).