The conceptual structure of traffic jams

Area-wide traffic jams develop through the propagation of queues from link to link, a process that resembles the growth of branches on a tree. The process is not well understood. In this paper, simple models for jam growth arising from a single bottleneck are developed for an idealized grid network. Under these idealized conditions, it has been shown that there are essentially two possible spatial configurations for a traffic jam on the type of network considered, each having a characteristic form sharing some of the properties of a fractal. More important, the models highlight an interesting dilemma in traffic management. A strategy that aims to minimize the rate of growth of a jam by a suitable allocation of queue storage space will actually encourage gridlock at the heart of the congested area, and conversely, a strategy that aims to defer gridlock will result in queues spread over a wider area. Extensive channelization (normally advocated in the interests of efficient traffic flow and safety) will also encourage longer queues. With hindsight, these conclusions seem obvious for any network whether imaginary or real, but they do not seem to have appeared in the literature, and the models give some indication of the size of the effects involved.