Paper detail

Braess' Paradox in a Generalised Traffic Network

The classical network configuration introduced by Braess in 1968 is of fundamental significance because Valiant and Roughgarden showed in 2006 that `the "global" behaviour of an equilibrium flow in a large random network is similar to that in Braess' original four-node example'. In this paper, a natural generalisation of Braess' network is introduced and conditions for the occurrence of Braess' paradox are formulated for the generalised network. The Braess' paradox has been studied mainly in the context of the classical problem introduced by Braess and his colleagues, assuming a certain type of networks. Specifically, two pairs of links in those networks are assumed to have the same volume-delay functions. The occurrence of Braess' paradox for this specific case of network symmetry was investigated by Pas and Principio in 1997. Such a symmetry is not common in real-life networks because the parameters of volume-delay functions are associated with roads physical and functional characteristics, which typically differ from one link to another (e.g. roads in networks are of different length). Our research provides an extension of previous studies on Braess' paradox by considering arbitrary volume-delay functions, i.e. symmetry properties are not assumed for any of the network's links and the occurrence of Braess' paradox is studied for a general configuration.