-1- Interactive Space-Time Diagram for Public Transit Ion Ho John C. Handley PARC, Inc., a Xerox Company 800 Phillips Road, MS 128-27E Webster, NY 14580 +1-585-265-8900, {Ion.Ho|John.Handley}@parc.com The use of visualization as an exploratory tool can help translate the massive operational data into something more digestible and structured for further analysis. We describe a system for visualizing quality of service data in public transit. Our visualization system has at its core a space time diagram (Marey graph), which is built on data from a commercial CAD/AVL system. The space time diagram depicts actual versus scheduled trajectories along fixed routes and allows immediate inspection of schedule adherence issues, bus bunching and more. Among our contributions to the state of the art are 1) data navigation tools that allow for filtering data by dates, routes, and directions; 2) mouse-over capability to show details of the vehicle including capacity, model, vehicle ID, and driver; 3) a Django Python-based Web framework which allows for rapid prototyping and extensibility. Keywords: visualization, Marey graph, D3, Django, Python, on time performance INTRODUCTION Many if not most transit agencies collect AVL (automated vehicle location) data and APC (automatic passenger count) data as part of their operations. Xerox Services offers such a system called OrbCAD™ which collects and manages that data in a SQL Server relational database. AVL data comprises GPS coordinate transmissions at intervals from 30s to 2 min depending on the system. Together with a schedule, one can determine the bus location relative to schedule and at what time the vehicle arrived at designated time points. In a space time diagram (as described below), one can see an entire day’s activity along a route at a glance. One can inspect schedule adherence, i.e., the time difference between when the bus is scheduled to arrive and when it actually arrives. One can look for instances of “bus bunching”, where buses, owing to delays, become too close to each other, causing one bus to pick up too many passengers while the other bus picks up the few remaining -- this introduces variability in the system that cascades into further delays because a crowded bus spends more time at a stop, loading and unloading passengers. APC data are used to estimate demand and load on the system (“ons” and “offs” at each stop). Intervals on a trip that have few or no passengers might be candidates for elimination, re-routing,