Origin–Destination Matrix Generation from Boarding–Alighting and Household Survey Data

Abstract

Approaches to demand estimation for public transport systems often use either primarily boarding–alighting data counts or travel survey data, based on a sampling of households. However, in the absence of a mechanism to capture transfer trips, boarding–alighting data have difficulty providing accurate demand estimation for large, complex transit systems. Travel survey data also present challenges in the estimation of total demand, when the demand is extrapolated from a limited sample of the population. The errors in this extrapolation often limit the estimation of demand to a coarse-grained zone-to-zone travel, as opposed to a fine-grained bus–rail stop to bus–rail stop. This paper presents the hybrid demand estimation (HDE) algorithm that synthesizes both types of data to produce an accurate, fine-grained demand estimation. Initially, HDE uses boarding–alighting data and a fluid flow model to estimate direct trips that match passenger boarding and alighting counts. Then, in an iterative process, three heuristics are used to introduce transfer trips and alter the set of no-transfer trips to create a better match to the household survey data. Both boarding–alighting and household travel survey data were collected for the Greater Mumbai, India, metropolitan area. The HDE algorithm was evaluated for Mumbai's multimodal suburban rail and bus system. In this network, the HDE algorithm produced significantly better matches to both types of data than other heuristic approaches. The demand estimation produced by HDE deviated from boarding–alighting and survey data by less than 1% across a range of metrics.