Scholarlink

<p>The advent of Industry 4.0 requires innovative approaches to ensure the production of high-quality goods within tight lead times. This paper delves into the application of cognitive architectures (CAs) in manufacturing, through the use of VSM-ACT-R 2, a model developed from the ACT-R architecture. VSM-ACT-R 2 enhances smart scheduling decisions that elevate productivity and maintain quality consistency. The model excels in four primary areas of manufacturing decision making: First, it implements tasks through decision-making algorithms and knowledge structures akin to those found in humans, supported by declarative memories that encapsulate intuitive and domain knowledge. Second, it reproduces decision-making processes at varying levels—from novice to expert—through production rules and retrieval systems that mimic human behavioral variations. Third, it models the learning trajectories of decision makers, governed by a control center that uses utility learning and reinforcement rewards. Last but not least, it incorporates metacognitive processes of reflection and evaluation of the progress of the selected approach through a dynamic reinforcement learning mechanism within a production system framework. We conclude by evaluation of this model, show the model learns how to give better suggestions for manufacturing solutions, and discuss its applications in using human-like decision-making cognitive model for manufacturing solutions, and its implications on integrating the model with Large Language Models for human-like decision-making alignment.</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

VSM-ACTR 2: a human-like decision making model with metacognition for manufacturing solutions

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月21日

查看全文
  • Siyu Wu,
  • Alessandro Oltramari,
  • Frank E. Ritter

摘要

The advent of Industry 4.0 requires innovative approaches to ensure the production of high-quality goods within tight lead times. This paper delves into the application of cognitive architectures (CAs) in manufacturing, through the use of VSM-ACT-R 2, a model developed from the ACT-R architecture. VSM-ACT-R 2 enhances smart scheduling decisions that elevate productivity and maintain quality consistency. The model excels in four primary areas of manufacturing decision making: First, it implements tasks through decision-making algorithms and knowledge structures akin to those found in humans, supported by declarative memories that encapsulate intuitive and domain knowledge. Second, it reproduces decision-making processes at varying levels—from novice to expert—through production rules and retrieval systems that mimic human behavioral variations. Third, it models the learning trajectories of decision makers, governed by a control center that uses utility learning and reinforcement rewards. Last but not least, it incorporates metacognitive processes of reflection and evaluation of the progress of the selected approach through a dynamic reinforcement learning mechanism within a production system framework. We conclude by evaluation of this model, show the model learns how to give better suggestions for manufacturing solutions, and discuss its applications in using human-like decision-making cognitive model for manufacturing solutions, and its implications on integrating the model with Large Language Models for human-like decision-making alignment.