Public transport routing

(outline, to be written out)

Public transport breaks the assumption that makes road routing simple: impedance is not a property of a link, but of a link at a moment in time. A connection that takes 20 minutes at 08:15 may be unavailable at 08:20 and take an hour at 22:00. Routing therefore has to be solved in three dimensions at once: departure time, transfers, and fares across operators.

Planned contents

• Time expanded graphs: representing a timetable as a graph whose nodes are combinations of time, stop and route, which turns a timetable problem back into a shortest path problem. Object Vision arrived at this representation independently of, and before, the published RAPTOR algorithm (Delling, Pajor & Werneck, 2015).
• Alternatives: time dependent graphs, RAPTOR and Connection Scan, and what each buys and costs
• Waiting, transfer and access time: the components of a public transport journey and their differing weights in generalised cost, see Travel impedance
• Departure time windows: why a single departure time is not an answer, and how to aggregate over a window using earliest arrival, the mean, or a profile
• Fares: non additive cost across operators, and why this cannot be modelled as link impedance
• Access and egress modes: walking or cycling to the stop, and combining them with the transit leg into a multimodal impedance
• GTFS as input: from timetable feed to routable network, see Create a routable network from GTFS files

Related

• GTFS · Travel impedance · Isochrones and catchment areas
• Dutch national network model, public transport