Self-Calibrated Transit Service Monitoring Using Automated Collected Data
This paper proposes a self-calibrated transit service monitoring framework that aims to obtain the performance of a transit system using automated collected data. We first introduce an event-based transit simulation model, which allows the detailed simulation of passenger travel behavior in a transit system, including boarding, alighting, and transfer walking. To estimate passenger path choices, we assume the path choices can be modeled using a C-logit model, and propose a simulation-based optimization model to estimate the path choice parameters based on automated fare collection and automated vehicle location data. The path choices can be estimated on a daily basis, which enables the simulation model to adapt to dynamic passenger behavior changes, and output more accurate network performance indicators for regular service monitoring such as train load, passenger travel time, and crowding at platforms. The proposed system eliminates the need for conventional monitoring equipment such as cameras at platforms and scaling/weighing systems on trains. The Hong Kong Mass Transit Railway (MTR) system is used as the case study. Results show that the model can well estimate the path choice behavior of passengers in the system. The output passenger exit flows are closer to the actual one compared to the two benchmark models (shortest path and uniform path choice).