Algorithmic Research Challenges Public Transit Routing Assumptions

A new study challenges the prevailing assumption that RAPTOR-based algorithms represent the state-of-the-art for public transit path-finding with unlimited transfers. The research, published as arXiv:2603.11729v2, systematically revisits classical Dijkstra-based approaches. It demonstrates that Time-Dependent Dijkstra (TD-Dijkstra) can outperform the MR algorithm. However, the study identifies a critical flaw in efficient TD-Dijkstra implementations: their reliance on filtering dominated connections during preprocessing. This filtering assumes passengers can always switch to a faster connection, a premise that proves unsound when stops incorporate buffer times. The core issue is the algorithm's inability to distinguish between seated passengers, who may continue their journey without waiting, and transferring passengers, who must respect the buffer time. This limitation arises because the preprocessing step cannot account for the different constraints faced by these two passenger states. The work addresses this gap by introducing a methodological correction to handle buffer times accurately within the algorithmic framework. The announcement type for the paper is listed as 'replace-cross'.