This repo contains the code of our paper Neural Combinatorial Optimization for Stochastic Flexible Job Shop Scheduling Problems, accepted at the 39th Annual AAAI Conference on Artificial Intelligence (AAAI-25).
The repo consists of 3 sub-repos. These sub-repos are used as follows:
The FJSP-DRL folder contains the main code of our project. Here, we apply the scenario processing module to the dual attention network to create SPM-DAN. This folder contains the code for DRL training and evaluation, as well as the code for the scenario processing module and the dual attention network. In addition, the code for the dispatching rules is also included in this folder.
The Job_Shop_Scheduling_Benchmark_Environments_and_Instances folder contains the code for the CP-SAT solver. It has both the deterministic CP-SAT formulation and our stochastic CP-stoch formulation. Both the code for creating schedules using CP-SAT and evaluating those schedules is included.
The L2D folder contains the code for our preliminary experiment in which we apply the scenario processing module to the L2D network.
Within each of these folders, there is a README file that explains the code in more detail.
To cite our work, please refer to the BibTeX below. This reference will be updated once the official AAAI proceedings are published.
@misc{smit2024neuralcombinatorialoptimizationstochastic,
title={Neural Combinatorial Optimization for Stochastic Flexible Job Shop Scheduling Problems},
author={Igor G. Smit and Yaoxin Wu and Pavel Troubil and Yingqian Zhang and Wim P.M. Nuijten},
year={2024},
eprint={2412.14052},
archivePrefix={arXiv},
primaryClass={cs.AI},
url={https://arxiv.org/abs/2412.14052},
}