Deep Reinforcement Learning-Based Optimization for Crew Allocation in Modular Building Prefabrication

Document Type

Conference Proceeding

Publication Date

1-1-2024

Abstract

Off-site construction has been steadily gaining popularity due to significant potentials and benefits in repeatability and standardization of processes. Since most of the construction activities are manually executed by labor force, it decreases productivity and manufacturing capacity. To this end, this study proposes a Deep Reinforcement Learning (DRL)-based decision-making tool for optimizing labor allocation to each ongoing activity with the aim to improve production pipelines by reducing delays and cost overruns. The study includes a framework that models the interdependencies between workstations and dependence on productivity of available labor and material resources to determine the optimal number of workers at various workstations, thereby reducing the make-span. The model utilizes the Q-learning algorithm along with real-time data on labor and material availability, current progress status, and deadlines of modules in the queue for predicting best possible action regarding distribution of workers to different workstations in the facility. The developed framework is elucidated through a case study to demonstrate the capabilities of the proposed model in automated handling and management of resources. The proposed framework contributes to the body of knowledge by (1) considering all renewable and non-renewable resources involved, (2) accounting for uncertainties, and (3) providing critical information required for well-informed decision-making in real time.

Publication Source (Journal or Book title)

Construction Research Congress 2024, CRC 2024

First Page

1317

Last Page

1326

This document is currently not available here.

Plum Print visual indicator of research metrics
PlumX Metrics
  • Usage
    • Abstract Views: 2
  • Captures
    • Readers: 7
see details

Share

COinS