Shortest Path with Alternatives for Uniform Arrival Times: Algorithms and Experiments

The Shortest Path with Alternatives (SPA) policy differs from classical shortest path routing in the following way: instead of providing an exact list of means of transportation to follow, this policy gives such a list for each stop, and the traveler is supposed to pick the first option from this list when waiting at some stop. First, we show that an optimal policy of this type can be computed in polynomial time for uniform arrival times under reasonable assumptions. A similar result was so far only known for Poisson arrival times, which are less realistic for frequency-based public transportation systems. Second, we experimentally evaluate such policies. In this context, our main finding is that SPA policies are surprisingly competitive compared to traditional shortest paths, and moreover yield a significant reduction of waiting times, and therefore improvement of user experience, compared to similar greedy approaches. Specifically, for roughly 25% of considered cases, we could decrease the expected waiting time by at least 20%. To run our experiments, we also describe a tool-chain to derive the necessary information from the popular GTFS-format, therefore allowing the application of SPA policies to a wide range of public transportation systems.