Multimodal routing with time-dependent taxi lifts

During the last decades, cities have spread widely, increasing the attraction of large metropolis toward nearby cities and peripheral zones. It led to a diversification of transportation offer in suburban areas in particular, with taxi like on-demand public transportation or new ride-haling services. However, most routing applications are not proposing to combine those rides with public transit and people are let with the burden of finding good combinations.
In this work, we focus on computing itineraries using taxi at the beginning or the end of the path to reach public transit, hence avoiding congestion in the city and travelling at a cheaper rate than a complete ride. However, building good itineraries combining taxi and public transit is expensive compare to walking or riding a bicycle as the number of stations that can be reached from a point is far more.
We propose a model that compares the inconvenience of taking a taxi to a number of transfers between trips depending on the taxi travel distance. In this setting, our resolution method computes Pareto optimal paths for earliest arrival time and profile queries with number of transfers as a second criterion. It extends the Trip Based Public Transit Routing algorithm [1]. An efficient pruning is also proposed to speed up the search. Tests are performed on the Korean network with historical traffic conditions.
[1] S. Witt. Trip-Based Public Transit Routing. In Algorithms – ESA 2015, Berlin, Heidelberg, 1025–1036, 2015