Passenger-to-train assignment model based on automated data

Abstract

This thesis aims at developing a methodology for assigning passengers to individual trains using: (i) fare transaction records from Automatic Fare Collection (AFC) system and (ii) the train tracking data from Automatic Train Regulation (ATR) system. The proposed Passenger-to-Train Assignment Model (PTAM) can provide a better understanding of capacity utilization and help assess the service quality in underground rail systems. PTAM is a probabilistic model that links each fare transaction to one (or multiple) feasible train itineraries. Key inputs to the model include the passenger walking speed distribution at stations. The thesis also develops methods to infer the parameters of the speed distribution using AFC and ATR data, while prior methods used either manually collected observations or statistically biased estimates. PTAM is applied in the context of Hong Kong's Mass Transit Railway (MTR) system and a series of applications are developed using PTAM output to assess the capacity utilization of the network, including trainload estimation, crowding assessment at stations, and animation of passenger movements in a playback mode.