# Cross-Problem Learning for Solving Vehicle Routing Problems

Source: [Cross-Problem Learning for Solving Vehicle Routing Problems](<https://www.sota2.com/research/paper/68fa73a81e209131cea5dbbd>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-26T02:54:13.197Z

Authors: Zhuoyi Lin, Yaoxin Wu, Bangjian Zhou, Zhiguang Cao, Wen Song, Yingqian Zhang, Senthilnath Jayavelu

Published: 2024-04-17T18:17:50.000Z

## Abstract

Existing neural heuristics often train a deep architecture from scratch for each specific vehicle routing problem (VRP), ignoring the transferable knowledge across different VRP variants. This paper proposes the cross-problem learning to assist heuristics training for different downstream VRP variants. Particularly, we modularize neural architectures for complex VRPs into 1) the backbone Transformer for tackling the travelling salesman problem (TSP), and 2) the additional lightweight modules for processing problem-specific features in complex VRPs. Accordingly, we propose to pre-train the backbone Transformer for TSP, and then apply it in the process of fine-tuning the Transformer models for each target VRP variant. On the one hand, we fully fine-tune the trained backbone Transformer and problem-specific modules simultaneously. On the other hand, we only fine-tune small adapter networks along with the modules, keeping the backbone Transformer still. Extensive experiments on typical VRPs substantiate that 1) the full fine-tuning achieves significantly better performance than the one trained from scratch, and 2) the adapter-based fine-tuning also delivers comparable performance while being notably parameter-efficient. Furthermore, we empirically demonstrate the favorable effect of our method in terms of cross-distribution application and versatility.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Prize-Collecting Traveling Salesperson Problem on PCTSP-20](<https://www.sota2.com/research/sota/prize-collecting-traveling-salesperson-problem-on-pctsp-20.md>): Prize Collecting Traveling Salesperson Problem; PCTSP-20; Optimality Gap: 1.02.

- [Orienteering Problem on OP-50](<https://www.sota2.com/research/sota/orienteering-problem-on-op-50.md>): Orienteering Problem; OP-50; Gap: 4.46.

- [Orienteering Problem on OP Average across scales 20 50 100](<https://www.sota2.com/research/sota/orienteering-problem-on-op-average-across-scales-20-50-100.md>): Orienteering Problem; OP Average across scales 20 50 100; Average Gap: 4.62.

- [Orienteering Problem on OP-20](<https://www.sota2.com/research/sota/orienteering-problem-on-op-20.md>): Orienteering Problem; OP-20; Optimality Gap: 0.85.

- [Prize-Collecting Traveling Salesman Problem on PCTSP Average across scales 20, 50, 100](<https://www.sota2.com/research/sota/prize-collecting-traveling-salesman-problem-on-pctsp-average-across-scales-20-50-100.md>): Prize-Collecting Traveling Salesman Problem; PCTSP Average across scales 20, 50, 100; Average Optimality Gap: 13.44.

- [Prize-Collecting Traveling Salesperson Problem on PCTSP-50 (Metrics: Cost, Gap)](<https://www.sota2.com/research/sota/prize-collecting-traveling-salesperson-problem-on-pctsp-50-metrics-cost-gap.md>): Prize Collecting Traveling Salesperson Problem; PCTSP-50; Optimality Gap: 14.11.

- [Prize-Collecting Traveling Salesperson Problem on PCTSP-100](<https://www.sota2.com/research/sota/prize-collecting-traveling-salesperson-problem-on-pctsp-100.md>): Prize Collecting Traveling Salesperson Problem; PCTSP-100; Optimality Gap: 25.19.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2404.11677>)

[PDF](<https://arxiv.org/pdf/2404.11677.pdf>)
