Optimization of a Deep Reinforcement Learning Policy for Construction Manufacturing Control

Ian Flood, Xiaoyan Zhou

2023

Abstract

The paper is concerned with the optimization of a deep learning approach for the intelligent control of a factory process that produces precast reinforced concrete components. The system is designed and optimized to deal with the unique challenges associated with controlling construction work, such as high customization of components and the need to produce work to order. A deep reinforcement learning strategy is described for training an artificial neural network to act as the factory control policy. The performance of the approach is maximized via a sensitivity analysis that ranges key modelling parameters such as the structure of the neural network and its inputs. This set of experiments is conducted on data acquired from a real factory. The study shows that the performance of the policy can be significantly improved by an appropriate selection of the modelling parameters. The paper concludes with suggestions for potential avenues for future research that could build upon the current work and further advance the approach.

Paper Citation

in Harvard Style

Flood I. and Zhou X. (2023). Optimization of a Deep Reinforcement Learning Policy for Construction Manufacturing Control. In Proceedings of the 13th International Conference on Simulation and Modeling Methodologies, Technologies and Applications - Volume 1: SIMULTECH; ISBN 978-989-758-668-2, SciTePress, pages 82-91. DOI: 10.5220/0012091400003546

in Bibtex Style

@conference{simultech23,
author={Ian Flood and Xiaoyan Zhou},
title={Optimization of a Deep Reinforcement Learning Policy for Construction Manufacturing Control},
booktitle={Proceedings of the 13th International Conference on Simulation and Modeling Methodologies, Technologies and Applications - Volume 1: SIMULTECH},
year={2023},
pages={82-91},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012091400003546},
isbn={978-989-758-668-2},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 13th International Conference on Simulation and Modeling Methodologies, Technologies and Applications - Volume 1: SIMULTECH
TI - Optimization of a Deep Reinforcement Learning Policy for Construction Manufacturing Control
SN - 978-989-758-668-2
AU - Flood I.
AU - Zhou X.
PY - 2023
SP - 82
EP - 91
DO - 10.5220/0012091400003546
PB - SciTePress