Project Summary
This project builds a novel data-driven, passenger-centric tool for BART scheduling and performance analysis. It has two components. The first develops statistical models that use BART hourly trip data and real-time vehicle ETA data to estimate passengers’ station-time-specific willingness to wait for the next train and to infer latent demand lost due to prolonged waiting times. Building on these estimates, the second component develops an efficient optimization framework based on the PI’s new queueing model for general on-off systems with delay-sensitive customers. The framework optimizes train schedules, including headways and dwell times, to maximize ridership while accounting for heterogeneous willingness to wait across stations and times. Preliminary April 2026 estimates for the Yellow Line suggest that 3.30% of total latent demand is lost due to long waiting times, including 2.47% on weekdays and 7.85% on weekends. Optimized schedules can recover much of this lost demand without adding vehicles. The proposed 12-month project will produce validated estimates of demand and passenger patience, an optimization prototype, scenario analysis results for BART operations, detailed schedule-change recommendations for BART staff, a practitioner-facing UC ITS report, and an academic journal article for submission to a top operations research outlet.
