Putting intelligence back into traffic lights

The first sets of traffic lights (in the 19th century) were controlled by a person stood beside them, however as technologies have developed, humans have increasingly been taken out of the control loop. Modern traffic lights are designed to monitor approaching vehicles automatically and this data then forms the input for algorithms to decide which direction should get a green light and for how long. Over time these algorithms have become more complex (to account for wider policy priorities such as pedestrian, cyclist and bus priority), but are limited by their inherent lack of flexibility in rapidly changing traffic conditions and high traffic flows, the very conditions where the adaptability of a human mind would be appropriate.This paper therefore reports on a study where teams of four human participants were given control of the traffic lights at four close proximity road junctions in a computer simulation. With no prior information, discussions or explanation of what the precise objective function of a control strategy should be, individuals were allowed to make up their own mind as to how they wanted the junction to operate and the consequential impacts on vehicles. Teams chose variously to either work cooperatively from the outset or to essentially ignore each other until other team members performance began to have a direct impact. Conclusions are presented illustrating the range of control strategies that emerged and their resulting impacts on the overall performance of the simulated traffic network.