Calibrating Rail Transit Assignment Models with Genetic Algorithm and Automated Fare Collection Data

Recently, automated fare collection (AFC) systems using smart card technology have become the main method for collecting urban rail transit (URT) fares in many cities around the world. Transaction data obtained through these AFC systems contain a large amount of archived information including how passengers use the URT system, and thus can be used in calibrating assignment models for precise passenger flow calculation. We present a methodology for calibrating URT assignment models using AFC data. The calibration approach uses a genetic algorithm-based framework with nonparametric statistical techniques. Three initial numerical tests show that the proposed approach finds more reasonable solutions than traditional approaches for the calibrated parameters. Furthermore, after calibration by the proposed approach, the existing assignment model delivers more accurate calculations of passenger flows in the network.