Data-driven planning of reliable itineraries in multi-modal transit networks

Multi-modal travel itineraries are based on traversing multiple legs using more than one mode of transportation. The more combinations of legs and modes, the more challenging it is for a traveler to identify a reliable itinerary. Transportation providers collect data that can increase transparency for reliable travel planning. However, this data has not been fully exploited yet, although it will likely form an important piece of future traveler information systems. Our paper takes an important step in this direction by analyzing and aggregating data from the operation of scheduled and unscheduled modes to create a reliability measure for multi-modal travel. We use a network search algorithm to evaluate itineraries that combine schedule-based long-distance travel with airlines with last-mile and first-mile drive times to efficiently identify the one with the highest reliability given a start time and travel-time budget. Our network search considers multiple origin and destination airports which impacts the first and last mile as well as the flight options. We use extensive historical datasets to create reliable itineraries and compare these with deterministic shortest travel-time itineraries. We investigate the amount of data that is required to create reliable multi-modal travel itineraries. Additionally, we highlight the benefits and costs of reliable travel itineraries and analyze their structure.