Deep Reinforcement Learning-based Ordering Mechanism for Performance Optimization in Multi-Echelon Supply Chains

Welcome to the repository for the Proximal Policy Optimization (PPO) model applied in the context of Supply Chain Ordering Management (SCOM). This project is a part of SCOM research and is accompanied by a paper titled "Deep Reinforcement Learning-based Ordering Mechanism for Performance Optimization in Multi-Echelon Supply Chains."

Overview

This repository contains the sample model code (drl_scm) used for developing and testing the ordering mechanism based on Proximal Policy Optimization. The model's purpose is to optimize ordering decisions in multi-echelon supply chains.

Key Features

• Model Code (drl_scm):

  • Use this sample model code for developing the reinforcement learning model for Supply Chain Ordering Management. Make necessary adjustments to adapt the code to testing environment.

• Sample Implementation (mod_test):

  • Illustrates a sample implementation of Experiments 2 and 3.
  • Highlights the observed results presented in Table 1 of corrections to the original paper.
  • Saves the order size decisions of each echelon as a data output CSV file in a designated folder location defined in the model code. Output_2 and Output_3 contain the output CSV files of the aforementioned two experiments respectively, and this output is used for deriving findings mentioned in Table A1 and B1 in the Appendix of the corrections to the original paper.

Citation

If you use the findings from the research work, please cite the original paper:

"Deep Reinforcement Learning-based Ordering Mechanism for Performance Optimization in Multi-Echelon Supply Chains."