Development of a Transfer-Cost-Based Passenger Flow Assignment Model for Beijing Rail Transit Network Using Automated Fare Collection Data

Literature review indicates that little is known about traveler behavior, such as transfer and route choices, in large transit systems because of the number of alternative routes involved and lack of empirical data. Even many transit route assignment models have been developed and ample automated fare collection (AFC) data has been collected, nearly no study has quantified how accurate of resulting flow assignments are, especially for transfer behaviors. However, as a multi-stakeholder system, it is essential to estimate passenger flows over Beijing rail transit network (BRTN) for revenue sharing and daily management/operation purpose. In this paper, major factors (including travel time and transfer cost) that influence passenger flow pattern in BRTN are considered. Specifically, a full transfer cost function, including transfer walking time, vehicle waiting time, and a penalty to additional transfers, is proposed to better simulate passenger's transfer behaviors. A general cost function for urban rail transit network is presented and corresponding route choice behaviour of travelers is analyzed. An improved logit-based model is then presented for solving this network flow assignment problem. The depth-first method is used to search for "effective paths" among all origin and destination (O-D) pairs. The average errors of estimated transfer flows from the proposed assignment model, which is proven to be more realistic in searching a set of reasonable paths, are below 20%. The results indicate that the models being developed in this study are capable of reasonably reproduce passenger's transfer and route choices and thus helpful for understanding the behaviours of rail transit passengers.