Deep-MARL-for-Dynamic-JSP - GitLite

KO | EN

gitlite — search

#javascript #python #react #typescript #llm #android #machine-learning #deep-learning #macos #java #awesome #cli #security #ios

Deep-MARL-for-Dynamic-JSP

(★ 98)

COR paper: A Deep Multi-Agent Reinforcement Learning Approach to Solve Dynamic Job Shop Scheduling Problem

File Explorer

Download Latest Version (.zip)

- decision_influence.xlsx
- exp_result_70.xlsx
- exp_result_80.xlsx
- exp_result_90.xlsx
- tour_result_70.xlsx
- tour_result_80.xlsx
- tour_result_90.xlsx
- Experiment_result.xlsx
- Validated_problem_instance_10^10.txt
- Validated_problem_instance_15^10.txt
- Validated_problem_instance_5^5.txt
- - event_breakdown_creation.py
  - event_job_creation.py
  - event_scenario_creation.py
- - static_fitness.py
  - static_genetic_algorithm.py
  - static_job_creation.py
  - static_spf.py
- __init__.py
- asset_machine.py
- brain_sequencing.py
- rule_sequencing.py
- validation.py
- saving_model.py
- validated.pt
- validated_10_70.pt
- validated_10_80.pt
- validated_10_90.pt
.gitignore
Main_experiment.py
Main_training.py
README.md
requirements.txt
Static_Experiment_ALL.py
Static_GA_run.py

# Installation Guide

1. Get the code

git clone https://github.com/RK0731/Deep-MARL-for-Dynamic-JSP

Downloads the entire project code from GitHub to your computer.

cd Deep-MARL-for-Dynamic-JSP

Moves into the project folder you just downloaded.

2. Python

Easy Recommended

Prerequisites

Git Needed to download the project code from GitHub.
Python 3 On Windows, be sure to check 'Add Python to PATH' during installation.

pip install -r requirements.txt

Installs the Python libraries listed in requirements.txt (or similar).

python <실행할 파일명>.py # README에서 정확한 실행 파일명을 확인하세요

Runs the Python script (or module).

✅ If it runs without errors and prints output in the terminal, it worked.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Introduciton

***Disclaimer on 2025 Mar 18:***

***Simulatior in this repository is vulnerable to job-overstay which leads to training instability and failure, and no longer maintained. Kindly consider using our latest sim model for your training and experiment: <https://github.com/RK0731/DRLforDynamicScheduling/tree/main/src>***

Repository for the code of COR paper entitled **A Deep Multi-Agent Reinforcement Learning Approach to Solve Dynamic Job Shop Scheduling Problem**

Please consider citing our paper if you found it helpful ^-^

@article{LIU2023106294,
    author = {Liu, Renke and Piplani, Rajesh and Toro, Carlos },
    title = {A deep multi-agent reinforcement learning approach to solve dynamic job shop scheduling problem},
    journal = {Computers & Operations Research},
    volume = {159},
    pages = {106294},
    year = {2023},
    issn = {0305-0548},
    doi = {https://doi.org/10.1016/j.cor.2023.106294},
    url = {https://www.sciencedirect.com/science/article/pii/S0305054823001582},
    }

## Repository Overview

This repo includes the code of simulation model, DRL learning algorithm, and sequencign rules for benchmark. Scripts' function can be identified by their prefix:
> 1. "asset_" under "src" directory: simulation model representing assets on shop floor, such as machines;
> 2. "brain_" under "src" directory: learning algorithm for job sequencing;
> 3. "event_" under "src/event_generator" directory: dynamic events on shop floor, such as job arrival;
> 4. "main_" in root directory: endpoints to start simulation processes, to train, or validate deep MARL-based algorithms;
> 5. "static_" under "src/static_experiment" directory: experiments on static problem instances.

Data and trained parameters can be found in folders:
> 1. "experiment_result": results of experiments in dynamic environments;
> 2. "GA_static_validation": static problem instances and experiment result on static problems;
> 3. "trained_models": the trained parameters.

## User Guide

To use our code as the benchmark, kindly refer to "trained_models" folder for the trained parameters, and use the class "network_value_based" within "brain_sequencing.py" file to build the neural network. The state and action functions can also be found in "Brain_sequencing.py" file.

An alternative way is to test your approach in our simulation model and context, you may create your algorithm and run the simulation in "Main_experiment.py" for comparison. Please refer to the comments in each module to see how to interact with the simulation model.