In previous scholarly literature, safety is understood as a main obstacle for introducing human–robot collaboration in industrial production. This interdisciplinary paper is concerned with the safety and regulation of human–robot collaboration and contribute to this debate through a case study of stakeholders in Sweden, exploring the views of the involved stakeholders which is largely absent in previous research literature. The case study concludes that while stressing some potential benefits, stakeholders within the industry are generally reluctant to human–robot collaboration. Current regulation and safety standards are understood to be one of the prominent obstacles to such solutions. Based on the perspectives of the stakeholders as well as an analysis of current regulation and safety standards, the paper identifies the following problems with current regulation: (i) existing categories and conceptualizations used to guide safety evaluation are problematic, (ii) intelligence and autonomous aspects of collaborative systems are not sufficiently addressed, (iii) current standards do not enable evaluation of the trade off between safety, efficiency and flexibility, and (iv) the regulation has a lack of focus on active safety and using the control system as a safety measure.

In an attempt to address these identified problems, the difference between traditional collaborative robots and intelligent human–robot collaboration is analyzed in the paper and a new safety approach is suggested, called Deliberative safety, which allows the humans and robots to switch between different safety measures based on the need for flexibility or efficiency to reach production goals. While considering system performance, we propose a taxonomy to better support the design of deliberative safety as well as five safety measures to use in a deliberative safety approach. These measures include available measures like perimeter safety, zone safety and reactive safety to more advanced measures like planned and active safety, and when used together, they can enable intelligent human–robot collaboration.

在以前的学术文献中，安全被认为是在工业生产中引入人机协作的主要障碍。这篇跨学科论文关注人机协作的安全性和监管，并通过对瑞典利益相关者的案例研究为这场辩论做出贡献，探索了相关利益相关者的观点，这在以前的研究文献中基本上没有。案例研究的结论是，虽然强调了一些潜在的好处，但行业内的利益相关者通常不愿意人机协作。当前的法规和安全标准被认为是此类解决方案的主要障碍之一。基于利益相关者的观点以及对现行法规和安全标准的分析，

为了解决这些已识别的问题，本文分析了传统协作机器人与智能人机协作之间的区别，并提出了一种新的安全方法，称为Deliberative safety，它允许人和机器人在不同的安全措施之间切换基于实现生产目标所需的灵活性或效率。在考虑系统性能的同时，我们提出了一种分类法，以更好地支持审慎安全的设计，以及在审慎安全方法中使用的五种安全措施。这些措施包括周边安全、区域安全和反应安全等可用措施，以及计划和主动安全等更高级的措施，当它们一起使用时，它们可以实现智能人机协作。