FMSG: Cyber: Resilient and Reliable Cyber-Physical-Human-Machine Teams: Toward Future of Cybermanufacturing

Industry 4.0旨在将传统的运营商控制系统转变为智能网络物理系统(CPS)。数字制造与设计创新研究院的首席技术官表示，“制造业产生的数据比任何其他经济部门都多”。大数据提供的竞争优势得到了广泛认可。然而，要实现这些好处，需要能够理解、建模和分析处理信息，并以一种有利于制造业知情决策的方式处理信息。机器人自动化的进步导致机器人在制造/组装/维修设施中与人类并肩工作。今天，缺乏安全保障阻碍了人与机器人的共生。此外，还缺乏能够有效利用制造设施中传感器提供的海量数据的高保真建模技术。未来制造种子奖助金(FMSG)网络制造项目旨在创造新的科学和培养新的人才，通过发展弹性和安全的人机协作，以及在存在外部干扰的情况下为制造系统中的部件流模型开发可靠的方法，来促进弹性和可靠的人-CPS系统的发展。我们将这些系统称为网络物理人机团队(CPHMT)。该项目的总体目标是开发一种集成的理论，用于制造系统的安全和高效的操作与网络物理人机团队(CPHMT)。部署CPHMT的主要困难之一是如何实现人机团队的弹性，并将该信息纳入系统级优化，以便在工厂车间进行决策。因此，本项目的总体目标是开发CPHMT弹性协调的安全方法，利用预测建模来估计安全指数，以用于构建高保真的制造部件流数学模型。具体地说，该项目1)开发弹性和安全的人机协作算法，使用可扩展和计算高效的心理过程计算模型，例如确定人的意图；2)开发有效、可靠和易于实现的方法来构建高保真的制造部件流数学模型，这是执行任何严格的、定量的分析和优化所必需的。3)应用：在实验室台架试验台上对理论结果进行验证，并通过工业案例进行实施。解决机器人控制问题的方法是利用深度学习的进展，利用操作员的运动模型、注意力、工作空间和可达性约束来预测人类的运动。然后将运动预测应用于一个耦合的动态运动模型，为机器人设计更安全的控制器。制造问题的解决方法是基于对具有CPHMT的制造系统中出现的随机过程的分析，并设计了基于人类运动意图的机器人操作的安全指数。作为结果，该项目展示了CPHMT方法的有效性，并为制造专业人员提供了有效的工具，用于生产操作和使用CPHMT的系统控制。本研究的成果为制造组织提供了一种新型的自动化--柔性自动化，使机器学习、人工智能能够快速适应不同产品的制造。为了实现它的部署，与机器学习、机器人和自动化相关的模块被开发为康涅狄格州大学新批准的机器人工程本科专业的一部分，其中描述和说明了CPHMT方法。这项研究提高了学生对机器学习、控制和制造的理解，以及他们解决STEM综合问题的能力。该项目得到了工程部(ENG)的土木工程、机械和制造业创新部(CMMI)、数学和物理科学部(MPS)的数学科学部(DMS)和社会、行为和经济科学部(SBE)的多学科活动办公室(SMA)的共同资助。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Industry 4.0 aims at converting traditional operator-controlled systems into smart cyber-physical systems (CPS). The chief technology officer of the Digital Manufacturing and Design Innovation Institute stated that “manufacturing generates more data than any other sector of the economy”. The competitive edge provided by the big data are widely acknowledged. However, achieving these benefits requires the ability to understand, model and analytically process information in a way that is conducive to informed decision-making in manufacturing. Advances in robotic automation has led to robots working alongside humans in manufacturing/assembly/repair facilities. Today, lack of safety assurance precludes human-robot symbiosis. Moreover, there is also a lack of high-fidelity modeling techniques that can effectively utilize the vast amount of data available from the sensors in manufacturing facilities. This Future Manufacturing Seed Grant (FMSG) CyberManufacturing project aims to create new science and develop new talent for the advancement of resilient and reliable human-CPS systems by developing resilient and safe coordination for human-machine teaming, and by developing reliable and robust methods for part flow models in manufacturing systems in the presence of external disturbances. We refer to these systems as cyber-physical human machine teams (CPHMT). The overarching goal of this project is to develop an integrated theory for safe and efficient operations of manufacturing systems with cyber-physical human machine teams (CPHMT).One of the main difficulties in deploying CPHMT is how to achieve resiliency of the human-machine teams and incorporate that information in the system level optimization for decision-making on the factory floor. Therefore, the overall goal of this project is to develop safety methods for resilient coordination of CPHMT, utilize predictive modeling to estimate safety index that can be used in construction of high-fidelity mathematical models of manufacturing parts flow. Specifically, the project 1) develops resilient and safe coordination algorithms for human-machine teaming using scalable and computationally efficient computational modeling of psychological processes such as determining human intentions, 2) develops effective, reliable, and easy-to-implement approach to construct high-fidelity mathematical models of manufacturing parts flow, which is necessary to perform any rigorous, quantitative analysis and optimization. 3) Application: Validate the theoretical results in lab bench-based testbed and implement them through industrial case studies. The approach to the robot control problems is based on utilizing advances in deep learning to predict the human motion using operator motion model, attention, and workspace and reachability constraints. Then the motion prediction is utilized in a coupled dynamic motion model to design safer controllers for robots. The approach to the manufacturing problems is based on analyses of random processes, which arise in manufacturing systems with CPHMT and designing a safety index for the robot to operate based on human motion intent. As an outcome, this project demonstrates the efficacy of the CPHMT approach and provide manufacturing professionals with effective tools for production operation and control of systems with CPHMT. The outcomes of this research provide manufacturing organizations with a novel type of automation - flexible automation, whereby the machine learning, artificial intelligence can be rapidly adapted to manufacture different products. To enable its deployment, modules related to machine learning, robotics and automation are developed to be offered as a part of newly approved Robotics engineering undergraduate major at UConn, where the CPHMT approach is described and illustrated. This study enhances the students' understanding of machine learning, control and manufacturing, and their capabilities to solve comprehensive STEM problems.This project is supported with co-funding from the Division of Civil, Mechanical and Manufacturing Innovation (CMMI) in the Engineering (ENG) Directorate, the Division of Mathematical Sciences (DMS) in the Directorate for Mathematical and Physical Sciences (MPS), and the Office of Multidisciplinary Activities (SMA) in the Directorate of Social, Behavioral and Economic Sciences (SBE).This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.