Exploring the Braess Paradox: Static Versus Dynamic Assignment

The Braess paradox is a well-known phenomenon initially observed in road traffic flow. It points out that increasing network capacity can lead to poorer performance in congested situations, when the drivers attempt to optimise their travel time individually. This paradox is not limited to road transport, but also extends to various information networks. In this article, we examine the Braess paradox in a closed network where demand remains constant. First, we determine the user and global optima of the deterministic system in stationary states with flow balancing. We present explicit formulae for the density intervals at which the Braess paradox occurs. We then compute Monte Carlo simulations of a stochastic mesoscopic traffic model using aggregate data obtained from a queueing model to explore the results. Static assignment models match the deterministic stationary results. In addition, the simulations assess the effectiveness of dynamic assignment, whereby drivers select routes in real time to minimise travel time. Interestingly, behaviour with dynamic assignment deviates from the generic static assignment results, particularly in highly congested situations. These results emphasise the significance of dynamic route selection in relation to Braess's paradox.