Dynamic passenger crowding and operations in rail transit systems: a validated framework of integrated data-driven agent-based simulation (AAAM)

Peak-hour crowding in rail transit especially in urban areas affects passengers' travel satisfaction and reduces operation performances for transit agencies. Interpreting dynamic performance such as the crowding patterns remain challenging due to passengers' time-varying movements and spatial-temporal changes. To increase the inference and decision-making capabilities of operations, this study develops an integrAted dAta-driven Agent-based siMulation (AAAM) to infer crowding effects and operational characteristics by combining smartcard and operational data. In the proposed framework, transit system characteristics, including topology, link connectivity, and vehicle capacities, are explicitly represented and automatically constructed from standard transit operation data. Each individual passenger's trip trajectory is simulated by modeling the sequence of activities, such as walking, waiting, onboard, and transferring, under the constraints of topology, capacity and operation time schedules. A benchmark case study is conducted with a full line of Beijing Subway incorporating smartcard data and operational records to illustrate the simulated dynamic conditions of the transit systems, including travel patterns, in-vehicle crowding, operational capacity utilization, and on-platform queuing. Validation and scalability assessment methods are developed to evaluate the consistency and functionality of the AAAM, demonstrating the framework's future capability for network-level assessments and large-scale cases. Results indicate that the majority of the differences between simulated value and data records are within 3 min. Optimization scenarios are further developed through the AAAM to calibrate operational characteristics such as peak-hour train occupancy and passengers' walking time. Results demonstrate the accuracy and practicality of the integrated data-driven agent-based simulation to capture behavioral and operational dynamics at robust, reliable, and scalable levels. • Develop an integrated data-driven agent-based simulation capturing individual and system dynamics. • Demonstrate the framework using a real-world network and corridor-level case study. • Analyze spatiotemporal patterns of crowding and congestion in rail transit operations. • Validate simulation results and examine scalability using individual smartcard records. • Calibrate effective operational parameters for peak and off-peak periods.