Simulating Trains

On the Internet, a lot of people have a lot of opinions on trains. Often, these result in discussions dominated by crayonistas making bold assertions about what is and isn’t possible and what is and isn’t a good idea.

This raises a natural question: is there a quantitative way to measure what is a good idea, based on publicly available data?

For a better sense of exactly what types of questions I have in mind, consider some of the recent project discussions in Los Angeles with Metro: the Crenshaw Line operating plan, the West Santa Ana Branch alignments, or the Crenshaw Line northern extension. In each of these projects, beyond the usual questions of politics and cost, there are considerations about how fast each line will be, the time cost of potential transfers for riders, and other nuanced operational questions.

The goal of this project, Capacity, is to provide an environment for implementing and evaluating the differences between operational alternatives. These evaluations should allow the user to answer questions such as: “If we pursued operating plan X, what would the median time spent waiting for transfers look like?” Or: “How would the median passenger wait time change if we ran twice as many trains on line Z?”

In order to provide such an environment, I’m going to attempt to use parallel discrete event simulation via ROSS to simulate the operations of transit system. The primary input will be GTFS exports, which provide information about the layout of a particular transit system, as well as prototypical schedules.

The primary caveat here is that I am neither a professional software developer nor a trained transit planner. I am, however, a systems and networking researcher with an interest in implementing efficient and equitable transit systems.