Scalable Algorithms for Bicriterion Trip-Based Transit Routing

TL;DR

This paper introduces scalable, partitioning-based extensions to Trip-Based Transit Routing, improving query times and preprocessing efficiency for large-scale transit networks through novel algorithms and multilevel graph partitioning techniques.

Contribution

It presents HypTBTR, a partitioning variant of TBTR, along with One-To-Many and multilevel partitioning methods to enhance transit routing performance and scalability.

Findings

HypTBTR reduces query times compared to TBTR.

One-To-Many algorithm accelerates profile queries and preprocessing.

Multilevel partitioning decreases fill-in computations significantly.

Abstract

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.