详情

原文传递 Estimating Transit Route OD Flows in the Presence of Multiple Latent Flow Patterns

题名：	Estimating Transit Route OD Flows in the Presence of Multiple Latent Flow Patterns
作者：	Mishalani, R. G.; McCord, M. R.; Goel, P.; Landgraf, A.
关键词：	Origin Destination (OD)##Transit routes##Automatic Passenger Count (APC)##Latent Flow Patterns##Education access##Empirical study##
摘要：	Route-level Origin-Destination (OD) flow matrices provide useful information for ridership forecasting, service planning (e.g., extending routes, splitting or combining routes, and introducing new routes), and control strategies development (e.g., short turning, expressing, and holding). Since directly observing OD flows via on-board surveys is time consuming and costly, many methodologies have been proposed to estimate route-level OD matrices from boarding and alighting counts (Ben-Akiva et al., 1985; Ben-Akiva, 1987; Kikuchi and Perincherry, 1992; Li and Cassidy, 2007; Li, 2009; Hazelton, 2010; Ji et al., 2014; Ji et al., 2015; Mishalani et al., 2017). Passenger boarding and alighting counts are relatively easier and less costly to collect than OD flow data. Moreover, many transit agencies are now collecting large quantities of boarding and alighting counts on a routine basis via Automatic Passenger Count (APC) technologies, thus, providing the opportunity to estimate up-to-date OD flows on an ongoing basis. While methods for estimating route-level OD matrices from boarding and alighting counts are based on different assumptions and employ different estimation approaches, they all assume, explicitly or implicitly, that the passenger flow patterns reflect a single underlying probability OD flow matrix. However, across the day and even within a time-of-day (e.g., morning, mid-day, and evening periods), different travelers are engaged in a variety of trip purposes, such as work, personal business, education, and shopping. Such variations within a time-of-day period are becoming even more prevalent with the advent of more complex life and work needs and duties across varying household structures. Therefore, passengers’ travel purposes, which are associated with different sets of origin-destination pairs and departure times within a time-of-day period would be more realistically represented by multiple underlying probability OD flows reflecting the variety of travel purposes. That is, it is conceivable that bus trips within a period could carry travelers that collectively exhibit different underlying probability OD flow patterns. It is possible to extend the formulations and methods proposed by Hazelton (2010) and Ji et al. (2015) to take into account the presence of multiple underlying OD flow matrices. However, the computational costs associated with doing so would render them infeasible. In contrast, the computationally efficient variational Bayes method (Mishalani et al., 2017) developed in a separate study offers the advantage of allowing for capturing the presence of multiple underlying OD flow matrices in a computationally feasible manner. In this study, this computationally efficient variational Bayes method is extended to capture the presence of multiple underlying OD flow matrices. In addition, a data-inspired simulation based evaluation is conducted to assess the value of recognizing multiple underlying probability OD flow matrices. Moreover, a preliminary empirical study is conducted to investigate the potential presence of multiple underlying probability OD flow matrices on an operational bus route.
总页数：	11
报告类型：	科技报告