Production Scheduling based on Deep Reinforcement Learning using Graph Convolutional Neural Network

Takanari Seito, Satoshi Munakata

Abstract

While meeting frequently changing market needs, manufacturers are faced with the challenge of planning production schedules that achieve high overall performance of the factory and fulfil the high fill rate constraint on shop floors. Considerable skill is required to perform the onerous task of formulating a dispatching rule that achieves both goals simultaneously. To create a useful rule independent of human expertise, deep reinforcement learning based on deep neural networks (DNN) can be employed. However, conventional DNNs cannot learn the important features needed for meeting both requirements because they are unable to process qualitative information included in these schedules, such as the order of operations in each resource and correspondence between allocated operations and resources. In this paper, we propose a new DNN model that can extract features from both numeric and nonnumeric information using a graph convolutional neural network (GCNN). This is done by applying schedules as directed graphs, where numeric and nonnumeric information are represented as attributes of nodes and directed edges, respectively. The GCNN transforms both types of information into the feature values by transmitting and convoluting the attributes of each component on a directed graph. Our investigation shows that the proposed model outperforms the conventional one.

Download


Paper Citation


in Harvard Style

Seito T. and Munakata S. (2020). Production Scheduling based on Deep Reinforcement Learning using Graph Convolutional Neural Network.In Proceedings of the 12th International Conference on Agents and Artificial Intelligence - Volume 2: ICAART, ISBN 978-989-758-395-7, pages 766-772. DOI: 10.5220/0009095207660772


in Bibtex Style

@conference{icaart20,
author={Takanari Seito and Satoshi Munakata},
title={Production Scheduling based on Deep Reinforcement Learning using Graph Convolutional Neural Network},
booktitle={Proceedings of the 12th International Conference on Agents and Artificial Intelligence - Volume 2: ICAART,},
year={2020},
pages={766-772},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0009095207660772},
isbn={978-989-758-395-7},
}


in EndNote Style

TY - CONF

JO - Proceedings of the 12th International Conference on Agents and Artificial Intelligence - Volume 2: ICAART,
TI - Production Scheduling based on Deep Reinforcement Learning using Graph Convolutional Neural Network
SN - 978-989-758-395-7
AU - Seito T.
AU - Munakata S.
PY - 2020
SP - 766
EP - 772
DO - 10.5220/0009095207660772