RL-Scheduling-2022-Summer

Reinforcement Learning approach to solve the Flexible Job Shop Scheduling Problem.

==============================

An optimized OpenAi Gym's environment to simulate the Flexible Job Shop Scheduling Problem.

Getting Started

This repository contains a environment for FJSSP as the JSSP file. We need to install the gym package first to use the environment.

pip install gym

Once installed and put JSSP file under the same folder as your main python file, the environment will be available in your OpenAi Gym environment and can be used to train a reinforcement learning agent:

import gym
import JSSP
env = gym.make("JSSP-v0", instance_path = INSTANCE_PATH)

Important: Your instance must follow section 2.6.1 of this paper.

How To Use

The observation provided by the environment contains both a boolean array indicating if the action is legal or not and the "real" observation

The observation is a tuple of self.state, and self.state is a dictionary of two entries:
job_machine_allocation : array of integers representing ith job's allocation, where -1 represents an empty allocation and -2 represents a finished job.
job_operation_status : array of integers representing ith job's current operation status.
For example,
observation = (-1,0,1,0,0,1)
self.state = {
job_machine_allocation: [-1,0,1]
job_operation_status: [0,0,1]
}
job 1 (operation 1) -> None
job 2 (operation 1) -> machine 1
job 3 (operation 2) -> machine 2

For each observation, the action space is spaces.Discrete(n) where n is the number of possible actions.

Project Organization

┌── README.md             <- The top-level README for developers using this project.
│
├── JSSP                  <- Contains the environment.
│
├── Instances             <- Contains many example instances of FJSSP.
│
├── GA                    <- Contains supporting files for genetic algorithm.
│
└── Solutions
    │
    ├── Solution_GA.ipynb                   <- The genetic algorithm for FJSSP.
    │
    ├── Solution_OR.ipynb                   <- The OR-tool algorithm for FJSSP.
    │
    ├── Solution_prepopulated_RL.ipynb      <- The prepopulated Q-table algorithm.
    │
    └── Solution_instance_division.ipynb    <- The instance division algorithm for FJSSP.

---

Credit

The README.md file presentation is inspired by this github repository, and the genetic algorithm in this repository is from this blog.

Contributors

Name: Hongjian Zhou
University: University of Oxford
Email: hongjian.zhou@exeter.ox.ac.uk

Name: Boyang Gu
University: Imperial College London
Email: boyang.gu19@imperial.ac.uk

Name: Chenghao Jin
Email: steven_jin_gbhg@foxmail.com

Publication

Zhou, Hongjian, Boyang Gu, and Chenghao Jin. "Reinforcement Learning Approach for Multi-Agent Flexible Scheduling Problems." arXiv preprint arXiv:2210.03674 (2022).

Link to Arxiv https://arxiv.org/abs/2210.03674