A Neural Column Generation Approach to the Vehicle Routing Problem with Two-Dimensional Loading and Last-In-First-Out Constraints

Yifan Xia,Xiangyi Zhang

2024-06-18

Abstract:The vehicle routing problem with two-dimensional loading constraints (2L-CVRP) and the last-in-first-out (LIFO) rule presents significant practical and algorithmic challenges. While numerous heuristic approaches have been proposed to address its complexity, stemming from two NP-hard problems: the vehicle routing problem (VRP) and the two-dimensional bin packing problem (2D-BPP), less attention has been paid to developing exact algorithms. Bridging this gap, this article presents an exact algorithm that integrates advanced machine learning techniques, specifically a novel combination of attention and recurrence mechanisms. This integration accelerates the state-of-the-art exact algorithm by a median of 29.79% across various problem instances. Moreover, the proposed algorithm successfully resolves an open instance in the standard test-bed, demonstrating significant improvements brought about by the incorporation of machine learning models. Code is available at <a class="link-external link-https" href="https://github.com/xyfffff/NCG-for-2L-CVRP" rel="external noopener nofollow">this https URL</a>.

Artificial Intelligence

What problem does this paper attempt to address?

This paper focuses on a variant of the Vehicle Routing Problem (VRP) called 2-Dimensional Loading Constraint (2L-CVRP) with Last-In-First-Out (LIFO) rule. VRP is a critical problem in the fields of logistics distribution, industrial warehousing, and on-demand delivery, with the objective of minimizing the total travel cost of the fleet to satisfy customer demands. 2L-CVRP introduces constraints on the two-dimensional loading of rectangular items and the loading order, making the problem more complex.

Most of the current research approximates the solution to 2L-CVRP using heuristic methods, with relatively fewer developments in exact algorithms. This paper proposes a novel algorithm called Neural Column Generation (NCG), which combines advanced machine learning techniques, particularly attention and recursion mechanisms, to accelerate the state-of-the-art exact algorithm. This approach improves the running speed of the algorithm on standard test instances, reducing the computational time by 29.79%, and successfully solves an unsolved instance in the standard test set.

The main contributions of this paper are as follows:
1. Proposal of a Neural Column Generation (NCG) algorithm that combines neural networks and column generation for the 2L-CVRP problem.
2. Development of a new machine learning model that integrates attention and recursion mechanisms to effectively handle the two-dimensional item loading problem with LIFO rule, achieving an accuracy rate of approximately 95% in predicting column feasibility.
3. Significant reduction in running time of the NCG algorithm on standard benchmark test instances, outperforming the existing best column generation algorithm by 29.79%.
4. Application of the NCG algorithm in Branch-and-Price-and-Cut (BPC) algorithm, which successfully solves open instances for the first time.

The paper also reviews existing non-learning methods and related work on neural column generation, pointing out that although previous machine learning methods have made breakthroughs in specific problems, they still face challenges in handling 2L-CVRP, particularly with LIFO rule. The proposed NCG approach effectively addresses this challenge by reducing the dependency on time-consuming feasibility checks traditionally used.