Detecting Braess Routes: an Algorithm Accounting for Queuing Delays With an Extended Graph
The Braess paradox is a counter-intuitive phenomenon whereby adding roads to a network results in higher travel time at equilibrium. In this paper we present an algorithm to detect the occurrence of this paradox in real-world networks with the help of an improved graph representation accounting for queues. The addition of queues to the network representation enables a closer match with real data. Moreover, we search for routes causing this phenomenon ("Braess routes") rather than links, and advocate removing such routes virtually from navigation systems so that the associated links can continue to serve other routes. Our algorithm relies on a convex optimization problem utilizing Beckmann potentials for road links as well as queues, and results in a route reconfiguration with reduced delay. We assume the availability of historical data to build the optimization model. We also assume the existence of a centralized navigation system to manage the routing options and remove the Braess routes. The theoretical solution demonstrates up to 12% delay reduction in a network from Montgomery County, Maryland. We validate the improvement with simulations.