A system of traffic lights designed to optimize flow during average conditions never works perfectly because the "average" never happens. So Researchers in Dresden, Germany are working on smarter sensors that react to actual, not estimated, traffic flow.
Even the more advanced traffic light systems in most cities are timed to work under conditions averaged from lots of data over the course of a day, so even if the system knows at 6 pm on a Friday to the signals based on estimated flow it won't recognize more people leaving 30 minutes early because of rain or held up because of a stalled car.
A team at the Dresden University of Technology are developing a sensor-based system that measures the number of vehicles going into an intersection and then leaving it. It takes this info and, using some algorithms, shares it with other traffic lights so they know to lengthen or shorten a light to get the group of cars through the intersection in a "green wave."
"It's a paradoxical effect that occurs in complex systems," said one researcher. "Surprisingly, delay processes can improve the system altogether. It is a slower-is-faster effect. You can increase the throughput - speed up the whole system - if you delay single processes within the system at the right time, for the right amount of time."
Initial tests show a reduction in delays of 56% for trams/buses, 9% for cars/trucks, and 36% for pedestrians. [Wired]