Learning to dispatch for Flexible Job Shop Scheduling via Deep Reinforcement Learning

This project adapts method proposed in this paper to solve Flexible Job Shop.

Installation

Anaconda

Create environment

conda env create -f conda_environment.yml

Then, activate environment

conda activate fjsp-rl

Reproduce experimental results

Run

python test_model.py

For problem with static and stochastic job arrival times, run

python test_model_stochastic.py

You can use Jupyter notebook to visualize the resulting schedule:

test_model.ipynb
test_model_stochastic.ipynb

Train new model

1. Go to params.py, change

config.size = 'MK01'        # this is the problem size, either MK01, MK02, ... MK10
config.device = 'cuda'      # device to train on, 'cpu' or 'cuda'
config.stochastic = False   # set this to True if you would like to train on problems with stochastic arrival time attributed to each job
config.progress_config.path_to_save_progress = ''   # path to save model weights, training and validation logs, directory of this path must be empty

2. Run

python train.py

Benchmark dataset

Dataset used for benchmark is available for download at:

https://people.idsia.ch/~monaldo/fjsp.html#Applications

Paper reference

@inproceedings{NEURIPS2020_11958dfe,
 author = {Zhang, Cong and Song, Wen and Cao, Zhiguang and Zhang, Jie and Tan, Puay Siew and Chi, Xu},
 booktitle = {Advances in Neural Information Processing Systems},
 editor = {H. Larochelle and M. Ranzato and R. Hadsell and M. F. Balcan and H. Lin},
 pages = {1621--1632},
 publisher = {Curran Associates, Inc.},
 title = {Learning to Dispatch for Job Shop Scheduling via Deep Reinforcement Learning},
 url = {https://proceedings.neurips.cc/paper/2020/file/11958dfee29b6709f48a9ba0387a2431-Paper.pdf},
 volume = {33},
 year = {2020}
}