• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>Simulation Innovation Workshop >A discrete event simulation-based multi-objective reinforcement learning reward function for optimizing manufacturing material handling
【24h】

A discrete event simulation-based multi-objective reinforcement learning reward function for optimizing manufacturing material handling

机译：基于离散的事件仿真的多目标强化学习奖励功能，用于优化制造材料处理

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Material handling is the activity of transporting materials from place to place within a manufacturing facility. The cost of material is significant and non-value-added for manufacturers; thus, managers strive to plan material handling so as to enable efficient production while minimizing cost. However, material handling is a very complex process, and consequently any algorithm or procedure to optimize material handling must consider multiple performance variables, including labor and equipment costs, time and distance, and the effects of possible material shortages. However, most conventional optimization procedures assume a single variable to be optimized. Therefore, combining or integrating multiple optimization variables into a single cost function is an active research area. In this paper we show how an abstract discrete event simulation can serve as the basis of an integrated cost function for a complex manufacturing material handling planning process. A weighted sum objective function applied to the simulation's results is the cost function. That cost function is used to calculate the reward in a larger implementation that applies reinforcement learning to improve material handling plans. The larger project is briefly described in this paper to set the integrated cost function in context. This may be the first work that shows how multiple material handling performance variables can be integrated into a single cost function suitable for optimization or machine learning.
机译：物料处理是将材料从一个地方运输到制造设施内的活性。材料的成本是制造商的显着和非增值;因此，管理人员努力计划材料处理，以便在最小化成本的同时实现高效的生产。然而，物料处理是一个非常复杂的过程，因此任何算法或用于优化材料处理的过程必须考虑多种性能变量，包括劳动力和设备成本，时间和距离，以及可能的材料短缺的影响。但是，大多数传统的优化过程假设要优化的单个变量。因此，将多种优化变量组合或集成到单个成本函数中是有源研究区域。在本文中，我们展示了抽象的离散事件仿真如何作为复杂制造材料处理规划过程的集成成本函数的基础。应用于模拟结果的加权和目标函数是成本函数。该成本函数用于计算较大的实现中的奖励，该实现应用加强学习以改善物料处理计划。本文简要介绍了较大的项目，以在上下文中设置综合成本函数。这可能是第一个工作，该工作显示如何集成多种材料处理性能变量如何集成到适合优化或机器学习的单一成本函数中。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号