[NeurIPS 2025] Learning to Insert for Constructive Neural Vehicle Routing Solver

This repository contains the code implementation of paper Learning to Insert for Constructive Neural Vehicle Routing Solver. In this paper, we propose Learning to Construct with Insertion-based Paradigm (L2C-Insert), a novel insertion-based learning framework for constructive NCO.

Dependencies

Python=3.8.6
matplotlib==3.5.2
numpy==1.23.3
pandas==1.5.1
pytz==2022.1
torch==1.12.1
torchaudio==0.12.1
torchvision==0.13.1
tqdm==4.64.1

Also see environment.yml.

Datasets and pre-trained models

The training and test datasets can be downloaded from Google Drive:

https://drive.google.com/drive/folders/1pJr3W8lbtAcbP9qfs82VJCjXbqOGQW02?usp=sharing

or Baidu Cloud:

https://pan.baidu.com/s/1H7bJJmS32-fgnXUgGFJYHw?pwd=7cgg

Implementation

Testing

TSP:
    cd L2C_Insert/TSP/Test
    python test_synthetic.py (on the synthetic dataset)
    python test_lib.py (on the tsplib dataset)
CVRP:
    cd L2C_Insert/CVRP/Test
    python test_synthetic.py (on the synthetic dataset)
    python test_lib.py (on the cvrplib dataset)

Training

TSP:
    cd /L2C_Insert/TSP/Train
    python train.py
CVRP:
    cd /L2C_Insert/CVRP/Train
    python train.py

Citation

If this repository is helpful for your research, please cite our paper:
"Fu Luo, Xi Lin, Mengyuan Zhong, Fei Liu, Zhenkun Wang, Jianyong Sun, and Qingfu Zhang, Learning to Insert for Constructive Neural Vehicle Routing Solver, The Thirty-ninth Annual Conference on Neural Information Processing Systems, NeurIPS 2025"

OR

@inproceedings{
luo2025learning,
title={Learning to Insert for Constructive Neural Vehicle Routing Solver},
author={Fu Luo and Xi Lin and Mengyuan Zhong and Fei Liu and Zhenkun Wang and Jianyong Sun and Qingfu Zhang},
booktitle={The Thirty-ninth Annual Conference on Neural Information Processing Systems},
year={2025},
url={https://openreview.net/forum?id=SXr3Dynctm}
}

---

Acknowledgements

• https://github.com/CIAM-Group/NCO_code/tree/main/single_objective/LEHD
• https://github.com/yd-kwon/POMO
• https://github.com/henry-yeh/GLOP