SCIMAI-Gym

Author Information

TITLE: SCIMAI-Gym
AUTHOR: Francesco Stranieri
INSTITUTION: University of Milano-Bicocca/Polytechnic of Turin
EMAIL: francesco.stranieri@polito.it

BibTeX Citation

If you use SCIMAI-Gym in a scientific publication, we would appreciate citations using the following format:

@article{Stranieri2024,
  title = {Combining deep reinforcement learning and multi-stage stochastic programming to address the supply chain inventory management problem},
  volume = {268},
  ISSN = {0925-5273},
  url = {http://dx.doi.org/10.1016/j.ijpe.2023.109099},
  DOI = {10.1016/j.ijpe.2023.109099},
  journal = {International Journal of Production Economics},
  publisher = {Elsevier BV},
  author = {Stranieri,  Francesco and Fadda,  Edoardo and Stella,  Fabio},
  year = {2024},
  month = feb,
  pages = {109099}
}

@misc{stranieri2022comparing,
  doi = {10.48550/ARXIV.2204.09603},
  url = {https://arxiv.org/abs/2204.09603},
  author = {Stranieri,  Francesco and Stella,  Fabio},
  keywords = {Machine Learning (cs.LG),  Artificial Intelligence (cs.AI),  Optimization and Control (math.OC),  FOS: Computer and information sciences,  FOS: Computer and information sciences,  FOS: Mathematics,  FOS: Mathematics,  68T07 (Primary),  90B06,  90B05 (Secondary)},
  title = {Comparing Deep Reinforcement Learning Algorithms in Two-Echelon Supply Chains},
  publisher = {arXiv},
  year = {2022},
  copyright = {Creative Commons Attribution 4.0 International}
}

Requirements

❗️ The following steps refer to the file ECML-PKDD_SCIMAI-Gym.ipynb.

To install and import necessary libraries, run the section:

Environment Setup

The code was tested with:

• Python 3.7
• Gym 0.19.0
• Ray 1.5.2
• Ax 0.2.1
• Matplotlib 3.4.3

Supply Chain Environment

To set up the Supply Chain Environment, run the section:

Reinforcement Learning Classes

📋 To change the configuration of the Supply Chain Environment (e.g., the number of product types, the number of distribution warehouses, costs, or capacities), edit the sub-section:

Supply Chain Environment Class

📋 To change the global parameters (e.g., the seed for reproducibility, the number of episodes for the simulations, or the directory to save plots), edit and run the section:

Global Parameters

Then, to initialize the Supply Chain Environment, run the section:

Supply Chain Environment Initialization

❗️ The output of this section will have the following format. Verify that the values are the same as the ones you defined.

--- SupplyChainEnvironment --- __init__
product_types_num is 1
distr_warehouses_num is 1
T is 25
d_max is [10]
d_var is [2]
sale_prices is [15]
production_costs is [5]
storage_capacities is [[5] [10]]
storage_costs is [[2] [1]]
transportation_costs is [[0.25]]
penalty_costs is [22.5]

Finally, to have some fundamental methods (e.g., the simulator or the plotting methods), run the section:

Methods

Baselines

To assess the DRL algorithms' performance, we established two different baselines. To initialize the Oracle and the (s, Q)-policy, run the sections:

Oracle
(s, Q)-Policy Class
(s, Q)-Policy Config [Ax]

📋 To change the (s, Q)-policy parameters (e.g., the total trials for the optimization or the number of episodes for each trial), edit the sub-section:

Parameters [Ax]

Finally, to have some fundamental methods (e.g., the methods for the Bayesian Optimization (BO) training or the plotting methods), run the section:

(s, Q)-Policy Methods [Ax]

Train BO Agent

To train the BO agent, run the section:

(s, Q)-Policy Optimize [Ax]

DRL Config

To change the DRL algorithms' parameters (e.g., the training episodes or the grace period for the ASHA scheduler), edit and run the sub-section:

Parameters [Tune]

📋 To change the DRL algorithms' hyperparameters (e.g., the neural network structure, the learning rate, or the batch size), edit and run the sub-sections:

Algorithms [Tune]
A3C Config [Tune]
PG Config [Tune]
PPO Config [Tune]

Finally, to have some fundamental methods (e.g., the methods for the DRL agents' training or the plotting methods), run the section:

Reinforcement Learning Methods [Tune]

Train DRL Agents

To train the DRL agents, run the section:

Reinforcement Learning Train Agents [Tune]

❗️ We upload the checkpoints of the best training instance for each approach and experiment, which can be used as a pre-trained model. For example, the checkpoint related to Exp 1 of the 1P3W scenario for the A3C algorithm is available at /Paper_Results/ECML-PKDD_2023_1P3W/1P3W/Exp_1/1P3W_2021-09-22_15-55-24/ray_results/A3C_2021-09-22_19-56-24/A3C_SupplyChain_2a2cf_00024_24_grad_clip=20.0,lr=0.001,fcnet_hiddens=[64, 64],rollout_fragment_length=100,train_batch_size=2000_2021-09-22_22-34-50/checkpoint_000286/checkpoint-286.

Results

To output the performance (in terms of cumulative profit) and the training time (in minutes) of the DRL algorithms, run the section:

Final Results

❗️ We save the plots of the best training instance for each approach and experiment. For example, the plots related to Exp 1 of the 1P3W scenario are available at /Paper_Results/ECML-PKDD_2023_1P3W/1P3W/Exp_1/1P3W_2021-09-22_15-55-24/plots.

The results obtained should be comparable with those in the paper. For example, for the 1P1W scenario, we achieve the following performance:

	A3C	PPO	VPG	BO	Oracle
Exp 1	870±67	1213±68	885±66	1226±71	1474±45
Exp 2	1066±94	1163±66	1100±77	1224±60	1289±68
Exp 3	−36±74	195±43	12±61	101±50	345±18
Exp 4	1317±60	1600±62	883±95	1633±39	2046±37
Exp 5	736±45	838±58	789±51	870±67	966±55