Scholarlink

<p>Global e-commerce growth forces traditional warehouses to shift from manual to fully automated operations. One popular way is&#xa0;to adopt Automated Storage and Retrieval System (AS/RS). However, AS/RS consumes a significant amount of energy&#xa0;to load/unload&#xa0;bulky and heavy items. One way to reduce AS/RS energy usage is by optimizing the crane scheduling, as it plays a crucial role in determining the sequence of tasks to be completed. This study proposes a novel dynamic&#xa0;scheduling for minimizing energy consumption while ensuring responsiveness based on Learning-based Simulation–Optimization (LSO) approach. LSO&#xa0;integrates Discrete-Event Simulation as environment and Deep Reinforcement Learning as agent which is optimized by Non-Dominated Sorting Genetic Algorithm II (NSGA-II)&#xa0;to cope with multi-objective problem. For the&#xa0;experiment, a single AS/RS was modeled virtually for training and testing under demand and supply uncertainties. The LSO approach is then&#xa0;compared with other well-known policies such as First Come First Serve, Nearest Neighbor, and Shortest Leg. The result shows that the LSO approach statistically outperforms other policies for both objectives, leading to greener and more responsive operations. Finally, managerial implications are provided where the LSO approach improves AS/RS utilization leading to more robust operation with the capability in providing real-time decisions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Dynamic crane scheduling for green automated warehousing: learning-based simulation-optimization approach

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年6月12日

查看全文
  • Zakka Ugih Rizqi,
  • Shuo-Yan Chou

摘要

Global e-commerce growth forces traditional warehouses to shift from manual to fully automated operations. One popular way is to adopt Automated Storage and Retrieval System (AS/RS). However, AS/RS consumes a significant amount of energy to load/unload bulky and heavy items. One way to reduce AS/RS energy usage is by optimizing the crane scheduling, as it plays a crucial role in determining the sequence of tasks to be completed. This study proposes a novel dynamic scheduling for minimizing energy consumption while ensuring responsiveness based on Learning-based Simulation–Optimization (LSO) approach. LSO integrates Discrete-Event Simulation as environment and Deep Reinforcement Learning as agent which is optimized by Non-Dominated Sorting Genetic Algorithm II (NSGA-II) to cope with multi-objective problem. For the experiment, a single AS/RS was modeled virtually for training and testing under demand and supply uncertainties. The LSO approach is then compared with other well-known policies such as First Come First Serve, Nearest Neighbor, and Shortest Leg. The result shows that the LSO approach statistically outperforms other policies for both objectives, leading to greener and more responsive operations. Finally, managerial implications are provided where the LSO approach improves AS/RS utilization leading to more robust operation with the capability in providing real-time decisions.