Learning to repeatedly solve routing problems

IF 1.6 4区计算机科学 Q4 COMPUTER SCIENCE, HARDWARE & ARCHITECTURE

Networks Pub Date : 2023-12-14 DOI:10.1002/net.22200

Mouad Morabit, Guy Desaulniers, Andrea Lodi

引用次数: 0

Abstract

In the last years, there has been a great interest in machine-learning-based heuristics for solving NP-hard combinatorial optimization problems. The developed methods have shown potential on many optimization problems. In this paper, we present a learned heuristic for the reoptimization of a problem after a minor change in its data. We focus on the case of the capacited vehicle routing problem with static clients (i.e., same client locations) and changed demands. Given the edges of an original solution, the goal is to predict and fix the ones that have a high chance of remaining in an optimal solution after a change of client demands. This partial prediction of the solution reduces the complexity of the problem and speeds up its resolution, while yielding a good quality solution. The proposed approach resulted in solutions with an optimality gap ranging from 0% to 1.7% on different benchmark instances within a reasonable computing time.

查看原文本刊更多论文

学会反复解决路由问题

在过去的几年里，人们对基于机器学习的启发式算法解决NP-hard组合优化问题产生了极大的兴趣。所开发的方法在许多优化问题上显示出潜力。在本文中，我们提出了一种学习启发式方法，用于在数据发生微小变化后对问题进行再优化。我们关注的是静态客户端(即相同的客户端位置)和需求变化情况下的可容车辆路径问题。给定原始解决方案的边缘，目标是在客户需求发生变化后，预测并修复最有可能保持在最优解决方案中的边缘。这种对解决方案的部分预测降低了问题的复杂性，加快了问题的解决速度，同时产生了高质量的解决方案。在合理的计算时间内，在不同的基准测试实例上，所提出的方法产生的解决方案的最优性差距从0%到1.7%不等。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Networks 工程技术-计算机：硬件

CiteScore

4.40

自引率

9.50%

发文量

审稿时长

12 months

期刊介绍： Network problems are pervasive in our modern technological society, as witnessed by our reliance on physical networks that provide power, communication, and transportation. As well, a number of processes can be modeled using logical networks, as in the scheduling of interdependent tasks, the dating of archaeological artifacts, or the compilation of subroutines comprising a large computer program. Networks provide a common framework for posing and studying problems that often have wider applicability than their originating context. The goal of this journal is to provide a central forum for the distribution of timely information about network problems, their design and mathematical analysis, as well as efficient algorithms for carrying out optimization on networks. The nonstandard modeling of diverse processes using networks and network concepts is also of interest. Consequently, the disciplines that are useful in studying networks are varied, including applied mathematics, operations research, computer science, discrete mathematics, and economics. Networks publishes material on the analytic modeling of problems using networks, the mathematical analysis of network problems, the design of computationally eﬃcient network algorithms, and innovative case studies of successful network applications. We do not typically publish works that fall in the realm of pure graph theory (without signiﬁcant algorithmic and modeling contributions) or papers that deal with engineering aspects of network design. Since the audience for this journal is then necessarily broad, articles that impact multiple application areas or that creatively use new or existing methodologies are especially appropriate. We seek to publish original, well-written research papers that make a substantive contribution to the knowledge base. In addition, tutorial and survey articles are welcomed. All manuscripts are carefully refereed.