Robotics and Computer-integrated Manufacturing最新文献_第5页

Learning-based robotic machining error prediction for high precision manufacturing 基于学习的高精度机器人加工误差预测

IF 11.4 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Robotics and Computer-integrated Manufacturing

Pub Date : 2026-01-02 DOI: 10.1016/j.rcim.2025.103217

Chaoyue Niu , Bin Chen , Simon Fletcher , Peace Onawumi , Erdem Ozturk , Mahdi Mahfouf , Visakan Kadirkamanathan

High precision machining with robots is an open challenge. Achieving precision of dimensional and geometrical features with robotic machining would require compensation via feedback control which relies on accurate error prediction. Machining error prediction is a complex problem in high-precision manufacturing, where effective solutions must accurately estimate geometrical errors in different workpieces while minimizing quality inspection costs. It is also compounded by the need for real-time estimation for feedback control. This paper introduces a novel approach for predicting the quality of milled workpieces using low-cost, in-process signals and machine learning. The proposed method fuses internal machine controller commands—comprising end-effector trajectory coordinates and angular changes of six revolute joints in the robotic arm—with external laser tracker sensing signals that capture the real trajectory of the milling tool and predicts dimensional errors as would be obtained by a Coordinate Measuring Machine (CMM). To overcome the lack of knowledge of the dependence of the part dimensional error on the available signals, models with varying combinations of the sensors and the length of the time window of historical data for inclusion in the model were evaluated. In addition, five machine learning algorithms were selected, trained, evaluated and validated on data from two distinct workpieces and various spatial configurations. The best machine learning model achieved a sevenfold improvement in dimensional error prediction compared to solely using laser tracker data, with mean absolute error reduced from 0.0756 mm to 0.0097 mm. This study demonstrates the feasibility of using low-cost, in-process sensing signals to predict high-precision quality dimensional data that is normally measured by costly CMMs, enabling rapid part quality inspection and significant potential cost reduction.

用机器人进行高精度加工是一个公开的挑战。机器人加工要实现尺寸和几何特征的精度，需要通过反馈控制进行补偿，而反馈控制依赖于精确的误差预测。加工误差预测是高精度制造中的一个复杂问题，有效的解决方案必须准确估计不同工件的几何误差，同时最大限度地降低质量检测成本。它还与反馈控制的实时估计需求相结合。本文介绍了一种利用低成本、过程中信号和机器学习预测铣削工件质量的新方法。该方法将机器内部控制器指令（包括末端执行器轨迹坐标和机械臂中六个旋转关节的角度变化）与外部激光跟踪器传感信号融合在一起，这些信号捕获铣刀的真实轨迹并预测由坐标测量机（CMM）获得的尺寸误差。为了克服零件尺寸误差对可用信号依赖性的缺乏，对具有不同传感器组合的模型和历史数据的时间窗口长度进行了评估。此外，在两种不同工件和不同空间配置的数据上选择、训练、评估和验证了五种机器学习算法。与仅使用激光跟踪器数据相比，最好的机器学习模型在尺寸误差预测方面实现了7倍的改进，平均绝对误差从0.0756 mm减少到0.0097 mm。本研究证明了使用低成本的过程传感信号来预测通常由昂贵的三坐标测量机测量的高精度质量尺寸数据的可行性，从而实现快速零件质量检测并显着降低潜在成本。

{"title":"Learning-based robotic machining error prediction for high precision manufacturing","authors":"Chaoyue Niu , Bin Chen , Simon Fletcher , Peace Onawumi , Erdem Ozturk , Mahdi Mahfouf , Visakan Kadirkamanathan","doi":"10.1016/j.rcim.2025.103217","DOIUrl":"10.1016/j.rcim.2025.103217","url":null,"abstract":"<div><div>High precision machining with robots is an open challenge. Achieving precision of dimensional and geometrical features with robotic machining would require compensation via feedback control which relies on accurate error prediction. Machining error prediction is a complex problem in high-precision manufacturing, where effective solutions must accurately estimate geometrical errors in different workpieces while minimizing quality inspection costs. It is also compounded by the need for real-time estimation for feedback control. This paper introduces a novel approach for predicting the quality of milled workpieces using low-cost, in-process signals and machine learning. The proposed method fuses internal machine controller commands—comprising end-effector trajectory coordinates and angular changes of six revolute joints in the robotic arm—with external laser tracker sensing signals that capture the real trajectory of the milling tool and predicts dimensional errors as would be obtained by a Coordinate Measuring Machine (CMM). To overcome the lack of knowledge of the dependence of the part dimensional error on the available signals, models with varying combinations of the sensors and the length of the time window of historical data for inclusion in the model were evaluated. In addition, five machine learning algorithms were selected, trained, evaluated and validated on data from two distinct workpieces and various spatial configurations. The best machine learning model achieved a sevenfold improvement in dimensional error prediction compared to solely using laser tracker data, with mean absolute error reduced from 0.0756 mm to 0.0097 mm. This study demonstrates the feasibility of using low-cost, in-process sensing signals to predict high-precision quality dimensional data that is normally measured by costly CMMs, enabling rapid part quality inspection and significant potential cost reduction.</div></div>","PeriodicalId":21452,"journal":{"name":"Robotics and Computer-integrated Manufacturing","volume":"100 ","pages":"Article 103217"},"PeriodicalIF":11.4,"publicationDate":"2026-01-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145886408","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A unified framework for large language model-guided reinforcement learning in digital twin industrial environments 数字孪生工业环境中大型语言模型引导强化学习的统一框架

IF 11.4 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Robotics and Computer-integrated Manufacturing

Pub Date : 2025-12-31 DOI: 10.1016/j.rcim.2025.103215

Haolin Fan , Edward Chow , Thomas Lu , Jerry Ying Hsi Fuh , Wen Feng Lu , Bingbing Li

Digital twin (DT) optimization in industrial environments faces persistent challenges, including sample inefficiency, extensive training requirements, and limited cross-domain adaptability. This paper presents a unified three-phase framework that integrates large language models (LLMs) with reinforcement learning (RL) via imitation learning (IL). The proposed approach comprises three key components: (1) offline expert demonstration collection using LLM-generated multi-agent coordination strategies, (2) offline and supervised IL to clone these strategies using a centralized training and decentralized execution (CTDE) architecture, and (3) lightweight RL fine-tuning to optimize the pre-trained policy. The system resolves equipment assignment conflicts and leverages coordination history for adaptive decision-making. Experiments in multi-agent industrial scenarios, including human–machine collaboration and fatigue-aware maintenance, demonstrate that our IL+RL hybrid reduces online training time by up to 96% while maintaining over 66% of optimal task performance, using only 4% of the training episodes required by standard RL. The approach also achieves 30%–40% task completion in zero-shot cross-domain settings (e.g., warehouse, manufacturing), and up to 99.7% with minimal fine-tuning. Conceptually, the framework establishes a new paradigm of ”language-conditioned IL,” where reasoning from general-purpose LLMs serves as an adaptive prior for efficient multi-agent coordination in DT. The results highlight how LLM-guided demonstrations can bridge symbolic reasoning and adaptive learning, offering both conceptual and practical advances for scalable, sample-efficient decision-making in Industry 5.0 systems.

工业环境中的数字孪生（DT）优化面临着持续的挑战，包括样本效率低下、广泛的培训要求和有限的跨领域适应性。本文提出了一个统一的三阶段框架，该框架通过模仿学习（IL）将大型语言模型（llm）与强化学习（RL）集成在一起。该方法包括三个关键部分：(1)使用llm生成的多智能体协调策略的离线专家演示集合，(2)使用集中训练和分散执行（CTDE）架构的离线和监督IL克隆这些策略，以及(3)轻量级RL微调以优化预训练策略。该系统解决了设备分配冲突，并利用协调历史进行自适应决策。在多智能体工业场景中的实验，包括人机协作和疲劳感知维护，表明我们的IL+RL混合方法将在线训练时间减少了96%，同时保持了超过66%的最佳任务性能，仅使用标准RL所需的4%的训练集。该方法还可以在零射击跨域设置（例如，仓库，制造）中实现30%-40%的任务完成率，并且在最小的微调下达到99.7%。从概念上讲，该框架建立了一个“语言条件IL”的新范式，其中来自通用llm的推理作为DT中有效的多代理协调的自适应先验。研究结果强调了法学硕士指导下的演示如何在符号推理和自适应学习之间架起桥梁，为工业5.0系统中可扩展的、样本高效的决策提供概念和实践上的进步。

{"title":"A unified framework for large language model-guided reinforcement learning in digital twin industrial environments","authors":"Haolin Fan , Edward Chow , Thomas Lu , Jerry Ying Hsi Fuh , Wen Feng Lu , Bingbing Li","doi":"10.1016/j.rcim.2025.103215","DOIUrl":"10.1016/j.rcim.2025.103215","url":null,"abstract":"<div><div>Digital twin (DT) optimization in industrial environments faces persistent challenges, including sample inefficiency, extensive training requirements, and limited cross-domain adaptability. This paper presents a unified three-phase framework that integrates large language models (LLMs) with reinforcement learning (RL) via imitation learning (IL). The proposed approach comprises three key components: (1) offline expert demonstration collection using LLM-generated multi-agent coordination strategies, (2) offline and supervised IL to clone these strategies using a centralized training and decentralized execution (CTDE) architecture, and (3) lightweight RL fine-tuning to optimize the pre-trained policy. The system resolves equipment assignment conflicts and leverages coordination history for adaptive decision-making. Experiments in multi-agent industrial scenarios, including human–machine collaboration and fatigue-aware maintenance, demonstrate that our IL+RL hybrid reduces online training time by up to 96% while maintaining over 66% of optimal task performance, using only 4% of the training episodes required by standard RL. The approach also achieves 30%–40% task completion in zero-shot cross-domain settings (e.g., warehouse, manufacturing), and up to 99.7% with minimal fine-tuning. Conceptually, the framework establishes a new paradigm of ”language-conditioned IL,” where reasoning from general-purpose LLMs serves as an adaptive prior for efficient multi-agent coordination in DT. The results highlight how LLM-guided demonstrations can bridge symbolic reasoning and adaptive learning, offering both conceptual and practical advances for scalable, sample-efficient decision-making in Industry 5.0 systems.</div></div>","PeriodicalId":21452,"journal":{"name":"Robotics and Computer-integrated Manufacturing","volume":"99 ","pages":"Article 103215"},"PeriodicalIF":11.4,"publicationDate":"2025-12-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145883839","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Constructing production equipment portraits and their realtime summarization for enabling automatic problem formulation in smart job scheduling 构建生产设备画像及其实时汇总，实现智能作业调度中问题的自动制定

IF 11.4 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Robotics and Computer-integrated Manufacturing

Pub Date : 2025-12-30 DOI: 10.1016/j.rcim.2025.103212

Jiaxiang Xie , Mingyuan Liu , Guofu Ding , Jian Zhang , Jianlin Fu , Haojie Chen

Disruptions in manufacturing workshops, such as new tasks or equipment breakdowns, frequently invalidate the current schedule, necessitating not only schedule adjustments but also the rapid redefinition of the scheduling problem itself. Although advanced scheduling frameworks have been developed for specific scheduling problems, a critical gap remains in a mechanism that can perform automatic problem formulation to respond to disruptions. Specifically, there is a lack of methods for automatically and effectively integrating complex and diverse equipment capability information into problem constraints. To address this, we propose a novel portrait-driven smart job scheduling paradigm. This paradigm introduces the perspective of a dynamic equipment capability portrait, which automatically abstracts raw equipment data into scheduling constraints. Therefore, this paradigm closes the loop from situation perception to problem definition, formulation, solving, and result interpretation. To support this novel paradigm, we propose a Hierarchical Production Equipment Capability Portrait (HPECP) method, which extends traditional self-portrait-based equipment portrait methods into task-driven portraits for decision-makers, thereby providing consistent and interpretable capability representations as well as constraints for the scheduling problem formulation. Additionally, to ensure its generalization and effectiveness in dynamic workshops, a technical implementation framework is proposed with four core components: standardized capability profile construction, task-capability relevance modeling, dynamic portrait reorganization, as well as a parameter-optimized strategy for the portrait reorganization process. At last, a simulation case derived from an actual workshop was performed to validate and analyze the effectiveness of the proposed paradigm. Analysis results demonstrated that the proposed HPECP method/technical framework provides an effective foundation for automatic problem formulation, thereby supporting future implementation and promotion of portrait-driven smart job scheduling in real workshops.

制造车间的中断，例如新的任务或设备故障，经常使当前的计划无效，不仅需要调整计划，而且需要快速重新定义计划问题本身。尽管针对特定的调度问题已经开发了先进的调度框架，但在能够执行自动问题制定以响应中断的机制方面仍然存在重大差距。具体而言，缺乏将复杂多样的设备能力信息自动有效地集成到问题约束中的方法。为了解决这个问题，我们提出了一种新的肖像驱动的智能作业调度范式。该范式引入了动态设备能力画像的视角，自动将原始设备数据抽象为调度约束。因此，这种范式封闭了从情境感知到问题定义、制定、解决和结果解释的循环。为了支持这种新范式，我们提出了一种分层生产设备能力画像（HPECP）方法，该方法将传统的基于自画像的设备画像方法扩展为决策者的任务驱动画像，从而为调度问题的制定提供一致和可解释的能力表示以及约束。此外，为了确保其在动态车间中的通用性和有效性，提出了一个技术实现框架，该框架包含四个核心组件：标准化能力轮廓构建、任务-能力关联建模、动态肖像重组以及肖像重组过程的参数优化策略。最后，通过一个实际车间的仿真案例，验证和分析了所提范式的有效性。分析结果表明，所提出的HPECP方法/技术框架为问题的自动制定提供了有效的基础，从而支持肖像驱动的智能作业调度未来在实际车间的实施和推广。

{"title":"Constructing production equipment portraits and their realtime summarization for enabling automatic problem formulation in smart job scheduling","authors":"Jiaxiang Xie , Mingyuan Liu , Guofu Ding , Jian Zhang , Jianlin Fu , Haojie Chen","doi":"10.1016/j.rcim.2025.103212","DOIUrl":"10.1016/j.rcim.2025.103212","url":null,"abstract":"<div><div>Disruptions in manufacturing workshops, such as new tasks or equipment breakdowns, frequently invalidate the current schedule, necessitating not only schedule adjustments but also the rapid redefinition of the scheduling problem itself. Although advanced scheduling frameworks have been developed for specific scheduling problems, a critical gap remains in a mechanism that can perform automatic problem formulation to respond to disruptions. Specifically, there is a lack of methods for automatically and effectively integrating complex and diverse equipment capability information into problem constraints. To address this, we propose a novel portrait-driven smart job scheduling paradigm. This paradigm introduces the perspective of a dynamic equipment capability portrait, which automatically abstracts raw equipment data into scheduling constraints. Therefore, this paradigm closes the loop from situation perception to problem definition, formulation, solving, and result interpretation. To support this novel paradigm, we propose a Hierarchical Production Equipment Capability Portrait (HPECP) method, which extends traditional self-portrait-based equipment portrait methods into task-driven portraits for decision-makers, thereby providing consistent and interpretable capability representations as well as constraints for the scheduling problem formulation. Additionally, to ensure its generalization and effectiveness in dynamic workshops, a technical implementation framework is proposed with four core components: standardized capability profile construction, task-capability relevance modeling, dynamic portrait reorganization, as well as a parameter-optimized strategy for the portrait reorganization process. At last, a simulation case derived from an actual workshop was performed to validate and analyze the effectiveness of the proposed paradigm. Analysis results demonstrated that the proposed HPECP method/technical framework provides an effective foundation for automatic problem formulation, thereby supporting future implementation and promotion of portrait-driven smart job scheduling in real workshops.</div></div>","PeriodicalId":21452,"journal":{"name":"Robotics and Computer-integrated Manufacturing","volume":"99 ","pages":"Article 103212"},"PeriodicalIF":11.4,"publicationDate":"2025-12-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145883837","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

3DprintMIND: An AI-Agent system using large language models and dynamic manufacturing knowledge graphs for smart manufacturing 3DprintMIND：使用大型语言模型和动态制造知识图的AI-Agent系统，用于智能制造

IF 11.4 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Robotics and Computer-integrated Manufacturing

Pub Date : 2025-12-30 DOI: 10.1016/j.rcim.2025.103214

Laiyi Li, Yongwen Zhang, Inno Lorren Désir Makanda, Pingyu Jiang

As an advanced manufacturing paradigm, 3D printing offers significant opportunities for smart manufacturing (SM). However, the prevalence of data silos in its production lines frequently hinders effective process analysis and decision-making. While large language models (LLMs) possess powerful analytical capabilities, their reliability in industrial scenarios is constrained by hallucinations and a disconnection from real-time operational data. Dynamic manufacturing knowledge graphs (MKGs), functioning as structured databases, can integrate structured and unstructured manufacturing data while providing LLMs with real-time data support. Meanwhile, the development of retrieval-augmented generation (RAG) and AI-Agent technologies provides a feasible pathway for the industrial application of LLMs. This study proposes an end-to-end framework ranging from knowledge integration to 3D printing production line applications to enable reliable LLM-driven SM. Firstly, leveraging the semantic analysis capabilities of LLMs, production data and manufacturing knowledge are integrated into a dynamic MKG. Subsequently, an advanced temporal graph network (TGN) model is developed for representation learning on the dynamic MKG, forming the retrieval foundation of an RAG system. A closed-loop manufacturing logic tailored to 3D printing production lines and a “semantic-to-structured” workflow for anomaly analysis were proposed. Finally, an AI-Agent system and a prototype software platform have been developed, and a case study on a laboratory 3D printing production line has been conducted to evaluate the TGN model’s performance and the AI-Agent system’s effectiveness. The results indicate that the dynamic MKG enables continuous learning for SM and provides industries with robust AI-driven data support. The TGN model significantly outperforms baseline models, yielding higher-quality dynamic embeddings for downstream knowledge retrieval tasks. Moreover, the AI-Agent system offers SM reliable intelligent analysis and decision support in 3D printing production lines.

作为一种先进的制造模式，3D打印为智能制造（SM）提供了重要的机会。然而，在其生产线中普遍存在的数据孤岛经常阻碍有效的过程分析和决策。虽然大型语言模型（llm）具有强大的分析能力，但它们在工业场景中的可靠性受到幻觉和与实时操作数据脱节的限制。动态制造知识图（MKGs）作为结构化数据库，可以集成结构化和非结构化制造数据，同时为llm提供实时数据支持。同时，检索增强生成（retrieval-augmented generation， RAG）和AI-Agent技术的发展为llm的产业化应用提供了可行的途径。本研究提出了一个从知识集成到3D打印生产线应用的端到端框架，以实现可靠的llm驱动的SM。首先，利用llm的语义分析能力，将生产数据和制造知识集成到动态MKG中。在此基础上，提出了一种用于动态MKG表示学习的高级时态图网络（TGN）模型，为RAG系统的检索奠定了基础。提出了适合3D打印生产线的闭环制造逻辑和异常分析的“从语义到结构”工作流。最后，开发了AI-Agent系统和原型软件平台，并在实验室3D打印生产线上进行了案例研究，以评估TGN模型的性能和AI-Agent系统的有效性。结果表明，动态MKG使SM能够持续学习，并为行业提供强大的人工智能驱动数据支持。TGN模型显著优于基线模型，为下游知识检索任务提供更高质量的动态嵌入。此外，AI-Agent系统为SM提供了可靠的3D打印生产线智能分析和决策支持。

{"title":"3DprintMIND: An AI-Agent system using large language models and dynamic manufacturing knowledge graphs for smart manufacturing","authors":"Laiyi Li, Yongwen Zhang, Inno Lorren Désir Makanda, Pingyu Jiang","doi":"10.1016/j.rcim.2025.103214","DOIUrl":"10.1016/j.rcim.2025.103214","url":null,"abstract":"<div><div>As an advanced manufacturing paradigm, 3D printing offers significant opportunities for smart manufacturing (SM). However, the prevalence of data silos in its production lines frequently hinders effective process analysis and decision-making. While large language models (LLMs) possess powerful analytical capabilities, their reliability in industrial scenarios is constrained by hallucinations and a disconnection from real-time operational data. Dynamic manufacturing knowledge graphs (MKGs), functioning as structured databases, can integrate structured and unstructured manufacturing data while providing LLMs with real-time data support. Meanwhile, the development of retrieval-augmented generation (RAG) and AI-Agent technologies provides a feasible pathway for the industrial application of LLMs. This study proposes an end-to-end framework ranging from knowledge integration to 3D printing production line applications to enable reliable LLM-driven SM. Firstly, leveraging the semantic analysis capabilities of LLMs, production data and manufacturing knowledge are integrated into a dynamic MKG. Subsequently, an advanced temporal graph network (TGN) model is developed for representation learning on the dynamic MKG, forming the retrieval foundation of an RAG system. A closed-loop manufacturing logic tailored to 3D printing production lines and a “semantic-to-structured” workflow for anomaly analysis were proposed. Finally, an AI-Agent system and a prototype software platform have been developed, and a case study on a laboratory 3D printing production line has been conducted to evaluate the TGN model’s performance and the AI-Agent system’s effectiveness. The results indicate that the dynamic MKG enables continuous learning for SM and provides industries with robust AI-driven data support. The TGN model significantly outperforms baseline models, yielding higher-quality dynamic embeddings for downstream knowledge retrieval tasks. Moreover, the AI-Agent system offers SM reliable intelligent analysis and decision support in 3D printing production lines.</div></div>","PeriodicalId":21452,"journal":{"name":"Robotics and Computer-integrated Manufacturing","volume":"99 ","pages":"Article 103214"},"PeriodicalIF":11.4,"publicationDate":"2025-12-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145883838","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Collaborative industrial product customization in real use environments through augmented reality: A case study of machine tool coolers 通过增强现实在真实使用环境中的协作工业产品定制：机床冷却器的案例研究

IF 11.4 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Robotics and Computer-integrated Manufacturing

Pub Date : 2025-12-26 DOI: 10.1016/j.rcim.2025.103216

Chih-Hsing Chu, Shau-Min Chen

Mass customization has become a business strategy for enabling companies to address individual customer needs at scale. Its technological realization requires design tools that enable effective engineering communication in the early design stages without the need for physical presence. This study introduces an integrated approach for collaborative industrial product customization conducted directly in real-use environments, using augmented reality (AR) technology as the interface. This study introduces an innovative augmented reality (AR)-based framework that enables customers to perform industrial product customization directly within their real-use environments, thereby transforming the conventional, time-consuming, and in-person customization process. Intelligent functions remotely accessible to on-site users through AR, such as streaming content control, ambient intelligence creation, parametric modeling, and in-situ functional simulation, facilitate the customization process via wireless networks. A real case of industrial cooler design demonstrates the feasibility of a prototyping tool implementing the proposed approach. A comparative study shows that the tool shortens customization steps that previously required hours or days to just minutes and eliminates repetitive iterations, achieving an overall efficiency improvement of over 80%. This work realizes the concept of conducting engineering design and simulation within in-situ environments through AR interfaces.

大规模定制已经成为一种商业策略，使公司能够大规模地满足个人客户的需求。它的技术实现需要设计工具，使有效的工程沟通在早期设计阶段，而不需要物理存在。本研究介绍了一种以增强现实（AR）技术为界面，直接在实际使用环境中进行协同工业产品定制的集成方法。本研究介绍了一种创新的基于增强现实（AR）的框架，该框架使客户能够直接在其实际使用环境中执行工业产品定制，从而改变了传统的、耗时的和面对面的定制过程。现场用户可以通过AR远程访问流媒体内容控制、环境智能创建、参数化建模、现场功能仿真等智能功能，通过无线网络方便定制过程。一个工业冷却器设计的实际案例证明了实现该方法的原型工具的可行性。一项比较研究表明，该工具将以前需要数小时或数天的定制步骤缩短到几分钟，并消除了重复的迭代，实现了80%以上的总体效率提高。本工作通过AR接口实现了在现场环境中进行工程设计和仿真的概念。

{"title":"Collaborative industrial product customization in real use environments through augmented reality: A case study of machine tool coolers","authors":"Chih-Hsing Chu, Shau-Min Chen","doi":"10.1016/j.rcim.2025.103216","DOIUrl":"10.1016/j.rcim.2025.103216","url":null,"abstract":"<div><div>Mass customization has become a business strategy for enabling companies to address individual customer needs at scale. Its technological realization requires design tools that enable effective engineering communication in the early design stages without the need for physical presence. This study introduces an integrated approach for collaborative industrial product customization conducted directly in real-use environments, using augmented reality (AR) technology as the interface. This study introduces an innovative augmented reality (AR)-based framework that enables customers to perform industrial product customization directly within their real-use environments, thereby transforming the conventional, time-consuming, and in-person customization process. Intelligent functions remotely accessible to on-site users through AR, such as streaming content control, ambient intelligence creation, parametric modeling, and in-situ functional simulation, facilitate the customization process via wireless networks. A real case of industrial cooler design demonstrates the feasibility of a prototyping tool implementing the proposed approach. A comparative study shows that the tool shortens customization steps that previously required hours or days to just minutes and eliminates repetitive iterations, achieving an overall efficiency improvement of over 80%. This work realizes the concept of conducting engineering design and simulation within in-situ environments through AR interfaces.</div></div>","PeriodicalId":21452,"journal":{"name":"Robotics and Computer-integrated Manufacturing","volume":"99 ","pages":"Article 103216"},"PeriodicalIF":11.4,"publicationDate":"2025-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145839890","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Graph-based multi-scale fusion learning for STEP-NC machining feature recognition 基于图的STEP-NC加工特征识别多尺度融合学习

IF 11.4 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Robotics and Computer-integrated Manufacturing

Pub Date : 2025-12-24 DOI: 10.1016/j.rcim.2025.103210

Zichuan Chai , Wenlei Xiao , Gang Zhao , Tianze Qiu , Yan Liu , Songyuan Xue , Oluwasheyi Oyename , Zheng Shi

The integration of AI into next-generation CAM systems has attracted significant research interest. Wherein, automatic feature recognition is a critical prerequisite before machining paths could be generated accordingly. Consequently, researchers have increasingly leveraged deep learning methodologies for geometric feature recognition from B-rep models. However, research targeting the recognition of machining features that ensure compatibility with downstream CAM toolpath generation remains limited. This paper proposes a multi-scale fusion graph neural network framework that embeds STEP-NC machining features to enhance their potency on the subsequent toolpath generation. Initially, feature semantics are extracted in accordance with the STEP-NC ISO 14649 standard, and a fusion network is constructed by integrating the adjacent-face aggregation of the GIN with the multi-head self-attention mechanism of the Graph Transformer. In the output layer, fine-grained label decomposition is performed based on standard definitions, enabling concurrent prediction of feature categories and their associated EXPRESS representations. Following pre-training, the model undergoes unsupervised fine-tuning on unlabeled real-world workpiece data to improve its generalization performance in practical manufacturing scenarios. Experimental results achieve over 85% recognition accuracy for real-part machining features in the automated manufacturing tasks.

将人工智能集成到下一代CAM系统中已经引起了极大的研究兴趣。其中，自动特征识别是加工轨迹生成的关键前提。因此，研究人员越来越多地利用深度学习方法从B-rep模型中识别几何特征。然而，针对加工特征的识别，以确保与下游凸轮刀具轨迹生成的兼容性的研究仍然有限。本文提出了一种嵌入STEP-NC加工特征的多尺度融合图神经网络框架，以增强其在后续刀具路径生成中的效力。首先，根据STEP-NC ISO 14649标准提取特征语义，并将GIN的邻接面聚合与Graph Transformer的多头自关注机制相结合，构建融合网络。在输出层中，基于标准定义执行细粒度标签分解，支持对特征类别及其相关EXPRESS表示进行并发预测。在预训练之后，该模型对未标记的真实工件数据进行无监督微调，以提高其在实际制造场景中的泛化性能。实验结果表明，在自动化制造任务中，该方法对实零件加工特征的识别准确率达到85%以上。

{"title":"Graph-based multi-scale fusion learning for STEP-NC machining feature recognition","authors":"Zichuan Chai , Wenlei Xiao , Gang Zhao , Tianze Qiu , Yan Liu , Songyuan Xue , Oluwasheyi Oyename , Zheng Shi","doi":"10.1016/j.rcim.2025.103210","DOIUrl":"10.1016/j.rcim.2025.103210","url":null,"abstract":"<div><div>The integration of AI into next-generation CAM systems has attracted significant research interest. Wherein, automatic feature recognition is a critical prerequisite before machining paths could be generated accordingly. Consequently, researchers have increasingly leveraged deep learning methodologies for geometric feature recognition from B-rep models. However, research targeting the recognition of machining features that ensure compatibility with downstream CAM toolpath generation remains limited. This paper proposes a multi-scale fusion graph neural network framework that embeds STEP-NC machining features to enhance their potency on the subsequent toolpath generation. Initially, feature semantics are extracted in accordance with the STEP-NC ISO 14649 standard, and a fusion network is constructed by integrating the adjacent-face aggregation of the GIN with the multi-head self-attention mechanism of the Graph Transformer. In the output layer, fine-grained label decomposition is performed based on standard definitions, enabling concurrent prediction of feature categories and their associated EXPRESS representations. Following pre-training, the model undergoes unsupervised fine-tuning on unlabeled real-world workpiece data to improve its generalization performance in practical manufacturing scenarios. Experimental results achieve over 85% recognition accuracy for real-part machining features in the automated manufacturing tasks.</div></div>","PeriodicalId":21452,"journal":{"name":"Robotics and Computer-integrated Manufacturing","volume":"99 ","pages":"Article 103210"},"PeriodicalIF":11.4,"publicationDate":"2025-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145823067","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A digital twin modeling framework with graphical software for rapid development of aircraft assembly systems 基于图形化软件的飞机装配系统快速开发数字孪生建模框架

IF 11.4 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Robotics and Computer-integrated Manufacturing

Pub Date : 2025-12-23 DOI: 10.1016/j.rcim.2025.103213

Ruihao Kang , Junshan Hu , Zhengping Li , Liangxiang Wang , Jincheng Yang , Wei Tian

Digital Twin (DT) technology is pushing manufacturing toward higher intelligence and adaptability. However, existing DT modeling methods still rely heavily on customization, lacking universality and scalability for assembly-oriented manufacturing systems. To address this limitation, this paper proposes a modular DT control framework that couples graphical interaction with reusable functional modules. Based on the classical five-dimensional DT model, the virtual entity is refined into geometric and physical models, and the service system is expanded into behavior and task models, enabling a clearer description and direct correspondence between system structure and operational logic. A behavior-oriented modeling workflow and a data-mapping mechanism are established to enhance scenario adaptability and reduce modeling effort. A graphical DT modeling platform is developed on top of this framework. Multiple robotic manufacturing prototypes, including robotic drilling, robotic gluing, and hybrid drilling systems, are constructed to assess the generality and reconfigurability of the proposed approach. A drilling experiment is performed on the robotic drilling system to validate the DT-based control execution mechanism. The resulting holes exhibit an average positioning error of 0.23 mm and a diameter error of 0.012 mm, both meeting aerospace drilling requirements. This confirms that virtual task commands can be accurately executed on physical system under the proposed DT framework. Overall, the DT prototype implementations and drilling experiment jointly verify the scalability of the framework and its DT-based control capability, providing a practical approach for the rapid development and deployment of DT prototypes in aircraft assembly systems.

数字孪生（DT）技术正在推动制造业向更高的智能和适应性发展。然而，现有的DT建模方法仍然严重依赖于定制，缺乏面向装配制造系统的通用性和可扩展性。为了解决这一限制，本文提出了一个模块化的DT控制框架，该框架将图形交互与可重用的功能模块相结合。在经典五维DT模型的基础上，将虚拟实体细化为几何和物理模型，将业务系统扩展为行为和任务模型，使系统结构与业务逻辑的描述更加清晰，直接对应。建立了面向行为的建模工作流和数据映射机制，增强了场景适应性，减少了建模工作量。在此框架的基础上开发了图形化DT建模平台。构建了多个机器人制造原型，包括机器人钻井、机器人粘合和混合钻井系统，以评估所提出方法的通用性和可重构性。在机器人钻井系统上进行了钻井实验，验证了基于dt的控制执行机制。所得到的孔的平均定位误差为0.23 mm，直径误差为0.012 mm，均满足航空航天钻井要求。这证实了在所提出的DT框架下，虚拟任务命令可以在物理系统上准确执行。总体而言，DT原型实现和钻井实验共同验证了框架的可扩展性及其基于DT的控制能力，为飞机装配系统中DT原型的快速开发和部署提供了实用方法。

{"title":"A digital twin modeling framework with graphical software for rapid development of aircraft assembly systems","authors":"Ruihao Kang , Junshan Hu , Zhengping Li , Liangxiang Wang , Jincheng Yang , Wei Tian","doi":"10.1016/j.rcim.2025.103213","DOIUrl":"10.1016/j.rcim.2025.103213","url":null,"abstract":"<div><div>Digital Twin (DT) technology is pushing manufacturing toward higher intelligence and adaptability. However, existing DT modeling methods still rely heavily on customization, lacking universality and scalability for assembly-oriented manufacturing systems. To address this limitation, this paper proposes a modular DT control framework that couples graphical interaction with reusable functional modules. Based on the classical five-dimensional DT model, the virtual entity is refined into geometric and physical models, and the service system is expanded into behavior and task models, enabling a clearer description and direct correspondence between system structure and operational logic. A behavior-oriented modeling workflow and a data-mapping mechanism are established to enhance scenario adaptability and reduce modeling effort. A graphical DT modeling platform is developed on top of this framework. Multiple robotic manufacturing prototypes, including robotic drilling, robotic gluing, and hybrid drilling systems, are constructed to assess the generality and reconfigurability of the proposed approach. A drilling experiment is performed on the robotic drilling system to validate the DT-based control execution mechanism. The resulting holes exhibit an average positioning error of 0.23 mm and a diameter error of 0.012 mm, both meeting aerospace drilling requirements. This confirms that virtual task commands can be accurately executed on physical system under the proposed DT framework. Overall, the DT prototype implementations and drilling experiment jointly verify the scalability of the framework and its DT-based control capability, providing a practical approach for the rapid development and deployment of DT prototypes in aircraft assembly systems.</div></div>","PeriodicalId":21452,"journal":{"name":"Robotics and Computer-integrated Manufacturing","volume":"99 ","pages":"Article 103213"},"PeriodicalIF":11.4,"publicationDate":"2025-12-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145823074","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Design, modeling and motion planning of a mobile continuum robot for in-situ inspection and maintenance in gas-insulated switchgear 气体绝缘开关柜现场检测维护移动连续机器人的设计、建模和运动规划

IF 11.4 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Robotics and Computer-integrated Manufacturing

Pub Date : 2025-12-22 DOI: 10.1016/j.rcim.2025.103205

Quan Xiao , Congjun Ma , Xuke Zhong , Yuqi Zhu , Xingxing You , Songyi Dian

In-situ inspection and maintenance of gas-insulated switchgear (GIS) are critical for ensuring power grid security and stable operation, as it can significantly reduce the current maintenance cycles which is extensive and costly due to the GIS disassembly and cleaning. However, navigating in/out via inspection ports to perform inspection and maintenance tasks in confined environments(e.g., arc-extinguishing chambers) is fairly challenging. This study proposes a novel multi-segment extra-slender (3 segments, diameter-to-length ratio

<

0.04) cable-driven mobile continuum robot (CDMCR) designed to enter confined spaces, execute maintenance tasks, and perform required joint configurations. The coupling between the mobile platform and the cable-driven continuum arm introduces significant redundancy. This redundancy complicates multi-constrained motion planning and reduces computational efficiency when exploring compact unstructured environments. To address this, we developed a real-time motion planner that incorporates mechanical configuration constraints, actuator limits, obstacle avoidance, and arc-surface constraints. The planner generates coordinated base and joint motions that track smooth end-effector trajectories. This enables global path planning from arbitrary initial states in prior-known scenes. Subsequently, an improved Follow-the-Leader (FTL) algorithm, inspired by the natural movement of snakes, ensures self-collision avoidance during end-path tracking. Laboratory and field evaluations demonstrate effective workspace coverage, comprehensive visual inspection capability within high-voltage GIS compartments, and robust success in solving random 6-DOF targets with responsive computation—validating both the robotic architecture and the proposed planning framework for practical power-equipment maintenance.

气体绝缘开关设备（GIS）的现场检测和维护对于确保电网的安全和稳定运行至关重要，因为它可以显着缩短当前由于拆卸和清洗GIS而产生的大量和昂贵的维护周期。但是，在密闭环境(例如：（灭弧室）是相当具有挑战性的。本研究提出了一种新型的多节超细长（3节，直径与长度比<；0.04）电缆驱动移动连续机器人（CDMCR），设计用于进入密闭空间，执行维护任务，并执行所需的关节配置。移动平台和电缆驱动连续臂之间的耦合引入了大量冗余。这种冗余使多约束运动规划变得复杂，并且在探索紧凑的非结构化环境时降低了计算效率。为了解决这个问题，我们开发了一个实时运动规划器，它结合了机械配置约束、执行器限制、避障和弧面约束。规划器生成协调的基座和关节运动，跟踪光滑的末端执行器轨迹。这使得全局路径规划从任意初始状态在先前已知的场景。随后，受蛇的自然运动启发，一种改进的Follow-the-Leader （FTL）算法确保在末端路径跟踪过程中避免自我碰撞。实验室和现场评估证明了有效的工作空间覆盖，在高压GIS隔间内的全面视觉检查能力，以及通过响应式计算解决随机6自由度目标的强大成功-验证了机器人架构和实际电力设备维护的拟议规划框架。

{"title":"Design, modeling and motion planning of a mobile continuum robot for in-situ inspection and maintenance in gas-insulated switchgear","authors":"Quan Xiao , Congjun Ma , Xuke Zhong , Yuqi Zhu , Xingxing You , Songyi Dian","doi":"10.1016/j.rcim.2025.103205","DOIUrl":"10.1016/j.rcim.2025.103205","url":null,"abstract":"<div><div>In-situ inspection and maintenance of gas-insulated switchgear (GIS) are critical for ensuring power grid security and stable operation, as it can significantly reduce the current maintenance cycles which is extensive and costly due to the GIS disassembly and cleaning. However, navigating in/out via inspection ports to perform inspection and maintenance tasks in confined environments(e.g., arc-extinguishing chambers) is fairly challenging. This study proposes a novel multi-segment extra-slender (3 segments, diameter-to-length ratio <span><math><mo><</mo></math></span>0.04) cable-driven mobile continuum robot (CDMCR) designed to enter confined spaces, execute maintenance tasks, and perform required joint configurations. The coupling between the mobile platform and the cable-driven continuum arm introduces significant redundancy. This redundancy complicates multi-constrained motion planning and reduces computational efficiency when exploring compact unstructured environments. To address this, we developed a real-time motion planner that incorporates mechanical configuration constraints, actuator limits, obstacle avoidance, and arc-surface constraints. The planner generates coordinated base and joint motions that track smooth end-effector trajectories. This enables global path planning from arbitrary initial states in prior-known scenes. Subsequently, an improved Follow-the-Leader (FTL) algorithm, inspired by the natural movement of snakes, ensures self-collision avoidance during end-path tracking. Laboratory and field evaluations demonstrate effective workspace coverage, comprehensive visual inspection capability within high-voltage GIS compartments, and robust success in solving random 6-DOF targets with responsive computation—validating both the robotic architecture and the proposed planning framework for practical power-equipment maintenance.</div></div>","PeriodicalId":21452,"journal":{"name":"Robotics and Computer-integrated Manufacturing","volume":"99 ","pages":"Article 103205"},"PeriodicalIF":11.4,"publicationDate":"2025-12-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145813786","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A mixed reality-assisted human-to-robot skill transfer approach for contact-rich assembly via visuomotor primitives 基于视觉运动原语的多接触装配的混合现实辅助人机技能转移方法

IF 11.4 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Robotics and Computer-integrated Manufacturing

Pub Date : 2025-12-22 DOI: 10.1016/j.rcim.2025.103208

Duidi Wu , Qianyou Zhao , Yuliang Shen , Junlai Li , Pai Zheng , Jin Qi , Jie Hu

Industrial assembly represents a core of modern manufacturing but poses significant challenges to the reliability and adaptability of robot systems. As manufacturing shifts toward intelligent production, there is an urgent need for efficient human-to-robot skill transfer methods for mutual cognition. However, current embodied intelligence research has primarily focused on household tasks, while human-level performance in dexterous and long-horizon tasks remains largely unexplored within real-world industrial applications. To bridge this gap, we propose a skill transfer framework and establish a contact-rich assembly benchmark. It integrates an MR-assisted digital twin system for low-cost and diverse demonstrations, an end-to-end generative visuomotor imitation learning policy for continuous action, and primitive skills covering industrially-inspired tasks such as peg insertion, gear meshing, and disassembly. Experiments across six tasks demonstrate high success rates and robust positional generalization. This study explores a novel pathway, it is hoped that it will provide valuable insights for future human–robot collaboration, and serve as a critical precursor for the integration of physical intelligence with generative AI. The project website is available at: https://h2r-mrsta.github.io/.

工业装配是现代制造业的核心，但对机器人系统的可靠性和适应性提出了重大挑战。随着制造业向智能生产的转变，迫切需要一种高效的人机相互认知的技能转移方法。然而，目前的具身智能研究主要集中在家庭任务上，而在现实世界的工业应用中，人类在灵巧和长期任务中的表现仍未得到充分的探索。为了弥补这一差距，我们提出了一个技能转移框架，并建立了一个富有接触的装配基准。它集成了磁共振辅助数字孪生系统，用于低成本和多样化的演示，端到端生成视觉运动模仿学习策略，用于连续动作，以及涵盖工业启发任务（如钉插入，齿轮啮合和拆卸）的原始技能。六个任务的实验证明了高成功率和稳健的位置泛化。本研究探索了一条新的途径，希望它将为未来的人机协作提供有价值的见解，并作为物理智能与生成式人工智能集成的重要先驱。该项目的网站是：https://h2r-mrsta.github.io/。

{"title":"A mixed reality-assisted human-to-robot skill transfer approach for contact-rich assembly via visuomotor primitives","authors":"Duidi Wu , Qianyou Zhao , Yuliang Shen , Junlai Li , Pai Zheng , Jin Qi , Jie Hu","doi":"10.1016/j.rcim.2025.103208","DOIUrl":"10.1016/j.rcim.2025.103208","url":null,"abstract":"<div><div>Industrial assembly represents a core of modern manufacturing but poses significant challenges to the reliability and adaptability of robot systems. As manufacturing shifts toward intelligent production, there is an urgent need for efficient human-to-robot skill transfer methods for mutual cognition. However, current embodied intelligence research has primarily focused on household tasks, while human-level performance in dexterous and long-horizon tasks remains largely unexplored within real-world industrial applications. To bridge this gap, we propose a skill transfer framework and establish a contact-rich assembly benchmark. It integrates an MR-assisted digital twin system for low-cost and diverse demonstrations, an end-to-end generative visuomotor imitation learning policy for continuous action, and primitive skills covering industrially-inspired tasks such as peg insertion, gear meshing, and disassembly. Experiments across six tasks demonstrate high success rates and robust positional generalization. This study explores a novel pathway, it is hoped that it will provide valuable insights for future human–robot collaboration, and serve as a critical precursor for the integration of physical intelligence with generative AI. The project website is available at: <span><span>https://h2r-mrsta.github.io/</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":21452,"journal":{"name":"Robotics and Computer-integrated Manufacturing","volume":"99 ","pages":"Article 103208"},"PeriodicalIF":11.4,"publicationDate":"2025-12-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145813784","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Knowledge graph-driven process reasoning of human-robot collaborative disassembly strategy for end-of-life products 知识图驱动的报废产品人机协同拆卸策略过程推理

IF 11.4 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Robotics and Computer-integrated Manufacturing

Pub Date : 2025-12-20 DOI: 10.1016/j.rcim.2025.103211

Jinhua Xiao , Zhiwen Zhang , Yu Zheng , Peng Wu , Sergio Terzi , Marco Macchi

Due to the complex structures and heterogeneous information inherent in End-of-Life (EOL) products, determining optimal disassembly solutions based on Human-Robot Collaboration (HRC) remains a challenging task. As structural and functional uncertainties in EOL products increase, traditional disassembly approaches struggle to meet the practical disassembly demands. Although various algorithms have been proposed for optimizing disassembly processes, significant challenges persist. These include the limited adaptability of existing models and difficulties in representing dynamic structured information effectively. To address these challenges, this study proposes a novel method combining knowledge graph-driven neural networks with an information decomposition module. This mechanism enables the network to discover structural semantic information and relational connections, facilitating the prediction of optimal disassembly strategies and enhancing the process reasoning capability of EOL product data and knowledge. Similarly, the proposed method provides reliable decision support for HRC disassembly task allocations and tool selections, enabling efficient and safe disassembly operations within complex disassembly processes. Finally, we demonstrate the method’s efficacy by using an example of an EOL battery pack, reasoning optimal disassembly strategies and potential process relations in the complex HRC disassembly scenario.

由于报废产品固有的复杂结构和异构信息，确定基于人机协作（HRC）的最佳拆卸方案仍然是一项具有挑战性的任务。随着EOL产品结构和功能不确定性的增加，传统的拆卸方法难以满足实际拆卸需求。尽管已经提出了各种算法来优化拆卸过程，但仍然存在重大挑战。这些问题包括现有模型的适应性有限，以及有效表示动态结构化信息的困难。为了解决这些挑战，本研究提出了一种将知识图驱动神经网络与信息分解模块相结合的新方法。该机制使网络能够发现结构语义信息和关系连接，便于预测最优拆卸策略，增强EOL产品数据和知识的过程推理能力。同样，该方法为HRC拆卸任务分配和工具选择提供了可靠的决策支持，实现了复杂拆卸过程中高效安全的拆卸操作。最后，我们以一个EOL电池组为例，证明了该方法的有效性，推理了复杂HRC拆卸场景下的最佳拆卸策略和潜在过程关系。

{"title":"Knowledge graph-driven process reasoning of human-robot collaborative disassembly strategy for end-of-life products","authors":"Jinhua Xiao , Zhiwen Zhang , Yu Zheng , Peng Wu , Sergio Terzi , Marco Macchi","doi":"10.1016/j.rcim.2025.103211","DOIUrl":"10.1016/j.rcim.2025.103211","url":null,"abstract":"<div><div>Due to the complex structures and heterogeneous information inherent in End-of-Life (EOL) products, determining optimal disassembly solutions based on Human-Robot Collaboration (HRC) remains a challenging task. As structural and functional uncertainties in EOL products increase, traditional disassembly approaches struggle to meet the practical disassembly demands. Although various algorithms have been proposed for optimizing disassembly processes, significant challenges persist. These include the limited adaptability of existing models and difficulties in representing dynamic structured information effectively. To address these challenges, this study proposes a novel method combining knowledge graph-driven neural networks with an information decomposition module. This mechanism enables the network to discover structural semantic information and relational connections, facilitating the prediction of optimal disassembly strategies and enhancing the process reasoning capability of EOL product data and knowledge. Similarly, the proposed method provides reliable decision support for HRC disassembly task allocations and tool selections, enabling efficient and safe disassembly operations within complex disassembly processes. Finally, we demonstrate the method’s efficacy by using an example of an EOL battery pack, reasoning optimal disassembly strategies and potential process relations in the complex HRC disassembly scenario.</div></div>","PeriodicalId":21452,"journal":{"name":"Robotics and Computer-integrated Manufacturing","volume":"99 ","pages":"Article 103211"},"PeriodicalIF":11.4,"publicationDate":"2025-12-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145784996","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0