Intriguingly Simple and Efficient Time-Dependent Routing in Road Networks

TL;DR

This paper introduces a simple, efficient algorithm for time-dependent routing in road networks that achieves near-optimal results by combining averaging travel times, shortest path computations, and restricted graph searches.

Contribution

The paper presents a novel, straightforward algorithm for time-dependent routing that avoids complex profile linking, offering near-optimal solutions with small subgraphs for efficient implementation.

Findings

Near-optimal results on benchmark instances

Reduced error with alternative route consideration

Small subgraphs enable efficient profile queries

Abstract

We study the earliest arrival problem in road networks with static time-dependent functions as arc weights. We propose and evaluate the following simple algorithm: (1) average the travel time in k time windows, (2) compute a shortest time-independent path within each window and mark the edges in these paths, and (3) compute a shortest time-dependent path in the original graph restricted to the marked edges. Our experimental evaluation shows that this simple algorithm yields near optimal results on well-established benchmark instances. We additionally demonstrate that the error can be further reduced by additionally considering alternative routes at the expense of more marked edges. Finally, we show that the achieved subgraphs are small enough to be able to efficiently implement profile queries using a simple sampling-based approach. A highlight of our introduced algorithms is that they do not rely on linking and merging profile functions.