Scalable Algorithms for Bicriterion Trip-Based Transit Routing

This paper proposes multiple extensions to the popular bicriterion transit routing approach -- Trip-Based Transit Routing (TBTR). Specifically, building on the premise of the HypRAPTOR algorithm, we first extend TBTR to its partitioning variant -- HypTBTR. However, the improvement in query times of HyTBTR over TBTR comes at the cost of increased preprocessing. To counter this issue, two new techniques are proposed -- a One-To-Many variant of TBTR and multilevel partitioning. Our One-To-Many algorithm can rapidly solve profile queries, which not only reduces the preprocessing time for HypTBTR, but can also aid other popular approaches such as HypRAPTOR. Next, we integrate a multilevel graph partitioning paradigm in HypTBTR and HypRAPTOR to reduce the fill-in computations. The efficacy of the proposed algorithms is extensively tested on real-world large-scale datasets. Additional analysis studying the effect of hypergraph partitioning tools (hMETIS, KaHyPar, and an integer program) along with different weighting schemes is also presented.

No reviews yet.

No comments yet.