Learning Large Neighborhood Search for Maritime Inventory Routing Optimization

TL;DR

This paper introduces a machine learning-enhanced local search method using graph neural networks to efficiently solve large-scale maritime inventory routing problems, outperforming traditional methods in quality and speed.

Contribution

It presents a novel integration of graph neural networks with local search for maritime routing, improving solution quality and computational efficiency.

Findings

Outperforms mixed-integer programming in solution quality and time

Effective neighborhood selection via graph neural networks

Scalable approach for large maritime routing instances

Abstract

Maritime inventory routing optimization is an important yet challenging combinatorial optimization problem. We propose a machine learning-based local search approach for finding feasible solutions of large-scale maritime inventory routing optimization problems. Given the combinatorial complexity of the problems, we integrate a graph neural network-based neighborhood selection method to enhance local search efficiency. Our approach enables a structured exploration of different neighborhoods by imitating an optimization-based expert neighborhood selection policy, improving solution quality while maintaining computational efficiency. Through extensive computational experiments on realistic instances, we demonstrate that our method outperforms direct mixed-integer programming as well as benchmark local search approaches in solution time and solution quality.