Improving traffic flow is not so much a matter of imposing particular behaviours as of creatingthe conditions under which the traffic can spontaneously organize itself in the most efficient manner. That philosophy doesn’t always sit easily with managers and planners, but it is looking increasingly to be the best way to approach all manner of social phenomena: to relinquish top-down control in favour of a faith in the bottom-up capacity of complex systems to find their own efficient modes of behaviour, given the opportunity.
This is illustrated by a reconsideration of how to coordinate traffic lights. The normal approach is to synchronize a periodic sequence of on-off times for aseries of lights at a cluster of intersections. But there is no reason to suppose either that the best sequence is strictly periodic or that it has toremain the same regardless of the traffic conditions. It turns out to be better to allow each individual traffic light to respond adaptively to the flow conditions at any moment.
The notion of allowing traffic signals individual autonomy could sound like a recipe for disaster– why should there be any guarantee that what is ‘best’ for one intersection will also suit what is happening at the others? However, if each signal issupplied with information not only about the traffic at that particularjunction it but also about the traffic coming from neighbouring junctions, thissharing of information can enable the system as a whole to find more effective, flexible solutions at any moment than are available from an insistence on regimented periodicity.
The idea is that traffic sensors placed a little before an intersection feed information about the incoming flow to each individual light-controlling system. This makes it possible to calculate the expected delays, and corresponding ‘trafficpressures’, at different parts of the network. Priority for green signals is then given to those parts experiencing the greatest pressures. In this way, the traffic itself controls the lights, rather than vice versa. Chance fluctuations,such as temporary lulls in the traffic on some routes, can be exploited torelieve congestion elsewhere. The behaviour that emerges can in fact look surprisingly synchronized, for example in the appearance of ‘waves’ of green lights that travel through the network.
Simulations of traffic flow on a network where autonomous lights are coordinated in this way show that overall average delays can be reduced by 30–40% relative to today’s state-of-the-art conventional control methods, and that the travel times for individual vehicles through the network actually become more predictable.
A simulationstudy of a real-world urban network – 13 intersections in the busy city centre of Dresden, complicated by tram lines and pedestrian crossings – has shown that self-organized traffic lights can significantly reduce the waiting times for all the modes of transport, including pedestrians. The same principle can be applied to other traffic-control measures such as speed limits: allowing them to adapt to the prevailing flow conditions can reduce congestion and delays.