Development of optimal operation algorithm for construction hoist based on artificial intelligence
연구지원기관: University of Michigan, Ripple, 연구기간: 2019.05 - 2021.07
The primary objective of this study is to minimize total transportation time—including both passenger waiting time and lifting time—by replacing experience-based manual hoist operation with a data-driven, AI-based decision-making framework. Through deep reinforcement learning, the project seeks to establish an adaptive control policy capable of outperforming conventional rule-based and human-operated hoist systems under diverse traffic patterns.[연구목표]
The primary objective of this study is to minimize total transportation time—including both passenger waiting time and lifting time—by replacing experience-based manual hoist operation with a data-driven, AI-based decision-making framework. Through deep reinforcement learning, the project seeks to establish an adaptive control policy capable of outperforming conventional rule-based and human-operated hoist systems under diverse traffic patterns.
[연구내용]
The project integrates a deep Q-network (DQN)–based control algorithm with a practical system architecture comprising wireless call devices, real-time hoist state sensors, a centralized control server, and automatic hoist controllers. A simulation-driven training environment and a mock-up validation platform are employed to evaluate performance under realistic construction traffic scenarios. The proposed approach is validated by comparing it against human operation and conventional dispatching algorithms using key performance indicators such as passenger waiting time, lifting time, and total transportation time.
[연구지원기관]
지원기관: 한국연구재단 (NRF), 학문후속세대양성사업
[연구기관]
Sep.2019 - Aug.2020
