The Fast Heuristic Algorithms and Post-Processing Techniques to Design Large and Low-Cost Communication Networks

TL;DR

This paper introduces two fast heuristic algorithms and post-processing techniques for designing large, low-cost communication networks by solving the NP-hard prize-collecting Steiner Tree Problem, demonstrating superior speed and solution quality on large benchmark instances.

Contribution

The paper presents novel fast heuristic algorithms and post-processing methods that significantly improve solution efficiency and quality for large-scale PCSTP instances in network design.

Findings

Algorithms solve large instances in 6-45 seconds.

Post-processing improves near-optimal solutions.

Algorithms outperform existing methods on benchmark instances.

Abstract

It is challenging to design large and low-cost communication networks. In this paper, we formulate this challenge as the prize-collecting Steiner Tree Problem (PCSTP). The objective is to minimize the costs of transmission routes and the disconnected monetary or informational profits. Initially, we note that the PCSTP is MAX SNP-hard. Then, we propose some post-processing techniques to improve suboptimal solutions to PCSTP. Based on these techniques, we propose two fast heuristic algorithms: the first one is a quasilinear time heuristic algorithm that is faster and consumes less memory than other algorithms; and the second one is an improvement of a stateof-the-art polynomial time heuristic algorithm that can find high-quality solutions at a speed that is only inferior to the first one. We demonstrate the competitiveness of our heuristic algorithms by comparing them with the state-of-the-art ones on the largest existing benchmark instances (169 800 vertices and 338 551 edges). Moreover, we generate new instances that are even larger (1 000 000 vertices and 10 000 000 edges) to further demonstrate their advantages in large networks. The state-ofthe-art algorithms are too slow to find high-quality solutions for instances of this size, whereas our new heuristic algorithms can do this in around 6 to 45s on a personal computer. Ultimately, we apply our post-processing techniques to update the bestknown solution for a notoriously difficult benchmark instance to show that they can improve near-optimal solutions to PCSTP. In conclusion, we demonstrate the usefulness of our heuristic algorithms and post-processing techniques for designing large and low-cost communication networks.