RouteFinder: Towards Foundation Models for Vehicle Routing Problems

2024-06-21

arXiv.org

AI Summary

This work addresses the challenge of poor generalization across the vast landscape of Vehicle Routing Problem (VRP) variants. We propose the first foundational model paradigm for VRP. Methodologically, we introduce a unified VRP environment capable of modeling diverse constraint combinations; design hybrid batch training, multi-variant reward normalization, and lightweight plug-and-play adapter layers; and integrate a Transformer encoder with global attribute embeddings to enable zero-shot transfer and efficient fine-tuning on unseen variants. Evaluated across 48 distinct VRP variants, our approach consistently outperforms existing learning-based methods, achieving substantial improvements in cross-variant generalization and training efficiency. The implementation is publicly available.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Computer Vision: Diffusion Models for Vision
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: User modeling for targeted and personalized online advertising

Abstract

This paper introduces RouteFinder, a comprehensive foundation model framework to tackle different Vehicle Routing Problem (VRP) variants. Our core idea is that a foundation model for VRPs should be able to represent variants by treating each as a subset of a generalized problem equipped with different attributes. We propose a unified VRP environment capable of efficiently handling any attribute combination. The RouteFinder model leverages a modern transformer-based encoder and global attribute embeddings to improve task representation. Additionally, we introduce two reinforcement learning techniques to enhance multi-task performance: mixed batch training, which enables training on different variants at once, and multi-variant reward normalization to balance different reward scales. Finally, we propose efficient adapter layers that enable fine-tuning for new variants with unseen attributes. Extensive experiments on 48 VRP variants show RouteFinder outperforms recent state-of-the-art learning methods. Code: https://github.com/ai4co/routefinder.

Problem

• Foundation model for Vehicle Routing Problems
• Unified environment for VRP variants
• Transformer-based encoder and reinforcement learning

Innovation

• Transformer-based encoder
• Global attribute embeddings
• Efficient adapter layers