Calibration of Rail Transit Assignment Model with Automated Fare Collection Data and a Parallel Genetic Algorithm

Abstract

In recent years, cities around the world have begun to use automated fare collection (AFC) systems with smart card technologies as the main method of collecting urban rail transit (URT) fares. Transaction data obtained through these AFC systems contain a large amount of archived information about how passengers use the URT system. These data can be used to calibrate assignment models for precise passenger flow calculations. However, this calibration typically is a computationally intensive problem because of multiroute searches, iteration strategies, and especially massive AFC data sets. This paper proposes a methodology for calibrating URT assignment models with AFC data and a parallel genetic algorithm. The calibration approach uses a framework based on a parallel genetic algorithm with nonparametric statistical techniques, which calibrate assignment model parameters by comparing observed and calculated travel time distributions. In initial case studies on the URT network in Beijing, the proposed approach found reasonable solutions for the calibrated parameters.