Locating inefficient links in a large-scale transportation network

Based on data from geographical information system (GIS) and daily commuting origin destination (OD) matrices, we estimated the distribution of traffic flow in the San Francisco road network and studied Braess’s paradox in a large-scale transportation network with realistic travel demand. We measured the variation of total travel time ΔT when a road segment is closed, and found that |ΔT| follows a power-law distribution if ΔT<0 or ΔT>0. This implies that most roads have a negligible effect on the efficiency of the road network, while the failure of a few crucial links would result in severe travel delays, and closure of a few inefficient links would counter-intuitively reduce travel costs considerably. Generating three theoretical networks, we discovered that the heterogeneously distributed travel demand may be the origin of the observed power-law distributions of |ΔT|. Finally, a genetic algorithm was used to pinpoint inefficient link clusters in the road network. We found that closing specific road clusters would further improve the transportation efficiency.