Application of System Theoretic Process Analysis to understand safety requirements for a domestic robot

家用机器人可以通过执行家务和为最终用户提供帮助和护理来提供显著的社会效益。然而，它们的部署可能会引入新的风险类型和/或导致意想不到的后果。为此，本项目研究了家用机器人在日常环境中的安全要求。因此，它属于EPSRC的人工智能和机器人研究领域。作为起点，该项目将应用系统理论过程分析（STPA）到家用机器人的应用。STPA是一种识别不安全控制行为和危险系统状态的技术，旨在生成详细的安全要求，以防止已识别的危险场景的发生。STPA是十多年前在麻省理工学院开发的。自成立以来，它已被采用在不同的行业，包括汽车行业。在过去，STPA已被证明在理解新产品的紧急行为和复杂相互作用方面非常有效，改进了以前的危害分析工具，如故障树分析（FTA），通过考虑故障发生的背景，识别由部件相互作用而不是单个部件故障引起的危害。系统理论事故模型和过程（STAMP）模型是支撑STPA的因果关系模型。STAMP采用分层安全控制结构，允许对过程行为（在本例中是家用机器人行为）执行不充分的安全约束，从而导致控制不充分，当应用于危险过程时，会导致危险的系统状态，从而导致事故发生。这与传统的危害分析工具（如FTA）不同，后者基本上认为事故是由于控制不足造成的，可以通过适当的设计要求来预防。这个项目的目的是加强家用机器人应用的安全案例。为此，我们的目标是通过将STPA应用于家用机器人系统，确定适用于家用机器人的高水平安全要求。应用场景可能侧重于家用机器人可能执行的具有代表性的任务，这些任务展示了诸如系统的基本能力需求，例如机器人导航、感知、操作等。在这方面，将利用STPA补充基于安全措施的风险评估。一个特别的重点将是机器人在开放式、动态环境中的操作，并额外努力展示部署在实际机器人系统上执行代表性任务的研究产品。研究方法的新颖性STPA方法应用于家用机器人的实例很少。自撰写本文以来，机器人领域已经取得了长足的进步，包括有腿机器人的发展，使其能够实现更广泛的应用场景。以前的STPA应用往往停留在高层次的抽象，没有深入研究具体的算法，并且缺乏在实际硬件上使用该方法的示例。该项目旨在解决这些限制，并探索如何将STPA和STAMP与机器人研究的新兴领域（如因果推理）结合起来。公司和合作者-主管：Dr. Lars Kunze -公司：Dyson Technology参考资料b[1] E. Mitka和S. G. Mouroutsos，“将印章系统安全工程方法应用于家用机器人的设计”，2015。

Research Context Domestic robots can offer significant societal benefits by performing household tasks and providing assistance and care to end-users. However, their deployment could introduce new types of risks and/or lead to unintended consequences. To this end, this project investigates the safety requirements for domestic robots in everyday environments. Hence, it falls within EPSRC's Artificial Intelligence and robotics research area. As a starting point, the project will apply System Theoretic Process Analysis (STPA) to domestic robot applications. STPA is a technique to identify unsafe control actions and hazardous system states with the aim to generate detailed safety requirements to prevent the occurrence of the identified hazardous scenarios. STPA was developed at MIT more than a decade ago. Since its inception, it has been adopted in different industries including the automotive sector. In the past, STPA has been shown to be very effective in understanding emergent behaviors and complex interactions of novel products, improving on previous hazard analysis tools such as Fault Tree Analysis (FTA) by considering the context in which faults occur, identifying hazards that arise from component interactions instead of just single component failures. The System-Theoretic Accident Model and Processes (STAMP) model is the causality model that underpins STPA. STAMP assumes a hierarchical safety control structure that allows inadequate enforcement of safety constraints on a processes behavior, in this case a domestic robots behavior, which leads to inadequate control which and when applied to the hazardous process results in a hazardous system state, which can lead to accidents occurring. This differs from conventional hazard analysis tools such as FTA, essentially saying accidents are caused by inadequate control, which can be prevented with suitable design requirements. Aims and Objectives The aim of this project is to strengthen the safety case for domestic robot applications. To do this, we aim to identify the high-level safety requirements that would apply to domestic robots through the application of STPA to domestic robot systems. Application scenarios might focus on representative tasks that a domestic robot might be expected to carry out, which demonstrate fundamental capability requirements of such as system, such as robot navigation, perception, manipulation, etc. In this context, STPA will be used to complement risk assessments based on safety measures. A particular focus will be on the operation of robots in open-ended, dynamic environments, with additional effort to show research products deployed on actual robot systems performing representative tasks. Novelty of the research methodologyThere have been few instances where the STPA method has been applied to domestic robots [1]. Since the time of writing, the field of robotics has seen considerable progress, including the development of legged robots that enable a wider range of applications scenarios. Previous applications of STPA tended to stay at a high level of abstraction, not delving into specific algorithms, and lacked examples of the method being used on real hardware. This project aims to address these limitations and explore how to combine STPA and STAMP with emerging areas of robotics research, such as causal inference. Companies and collaborators - Supervisor: Dr. Lars Kunze - Companies: Dyson Technology References [1] E. Mitka and S. G. Mouroutsos, "Applying the stamp system safety engineering methodology to the design of a domestic robot," 2015.