Data-Driven Transit Network Design at Scale

Abstract

Mass transit remains the most efficient way to service a densely packed commuter population. However, reliability issues and increasing competition in the transportation space have led to declining ridership across the United States, and transit agencies must also operate under tight budget constraints. Recent attempts at using bus network redesign to improve ridership have attracted attention from various transit authorities. However, the analysis seems to rely on ad hoc methods, for example, considering each line in isolation and using manual incremental adjustments with backtracking. We provide a holistic approach to designing a transit network using column generation. Our approach scales to hundreds of stops, and we demonstrate its usefulness on a case study with real data from Boston.