/mta-gtfs-rl-irl

This project explores the application of Reinforcement Learning (RL) and Inverse Reinforcement Learning (IRL) to optimize bus transit schedules using real-world transit data. The primary objective is to develop a predictive model that can generate efficient bus schedules, reduce delays, and improve the reliability of public transportation services.

Primary LanguageJupyter Notebook

mta-gtfs-rl-irl

This project explores the application of Reinforcement Learning (RL) and Inverse Reinforcement Learning (IRL) to optimize bus transit schedules using real-world transit data. The primary objective is to develop a predictive model that can generate efficient bus schedules, reduce delays, and improve the reliability of public transportation services. By combining RL, which learns optimal actions based on rewards, with IRL, which derives rewards from observed behavior, this model aims to achieve a balance between planned schedules and realistic, adaptive operations based on historical patterns. The project utilizes data from the New York City MTA-GTFS bus line, modeling bus behavior and transit flow to simulate and predict optimal routes.
Dataset: https://s3.amazonaws.com/MTABusTime/B63-2011-04-03_2011-05-03.zip