Carpooling with Transfers and Travel Time Uncertainty

Abstract

Carpooling is known to have lower CO2 emissions compared to driving individually. One of the limitations of carpooling is that matched drivers and passengers need to have similar itineraries, or their generalized costs will be high. By allowing a single transfer at a designated transfer hub, their itineraries need to be only partially similar. We allow for transfers within the carpooling system and between carpooling and public transport. Thereby, we include travel time uncertainty to evaluate its effect on carpooling with transfers. We model the ride-matching problem with transfers and travel time uncertainty as a two-stage stochastic programming problem. The results indicate that a single transfer hub can already reduce the average generalized cost of passengers by 15%. When travel times are uncertain, commuters tend to find a match that performs relatively well in every traffic situation, rather than one that performs well for only one scenario.