A spatial algorithm for the analysis of transportation systems using statistical model checking

TL;DR

This paper introduces an automated, data-driven methodology that constructs probabilistic models of transportation systems from vehicle location data, enabling performance evaluation and system modification analysis through simulation and statistical model checking.

Contribution

It presents a novel, largely route-agnostic approach that infers transportation system models directly from data using automated map generation and statistical analysis.

Findings

Effective evaluation of service performance using probabilistic models.

Insights into the impact of scheduling strategies on bus service performance.

Models validated through case studies on Edinburgh and Seattle transit systems.

Abstract

We present an automated methodology for using Automatic Vehicle Location measurements of public transportation vehicles to construct a probabilistic model. The model not only allows for accurate evaluation of service performance, but also makes it possible to study the effects of system modifications a priori. The methodology is almost entirely agnostic to otherwise important details of the service -- in particular its route and the location of stops. Instead, it infers this from the data using automated map generation techniques. The behaviour of vehicles in the model is analysed using computer simulation combined with statistical model checking. We present two case studies involving the Airlink service in Edinburgh and the Bellevue Express in Seattle. To demonstrate the usefulness of the approach, we analyse the impact of the scheduling strategies of bus holding and speed modification on the Airlink's performance. The data and code used to create the figures are publicly available online.