School of Computer Science and Technology, Harbin Institute of Technology, Harbin 150001, China.
LIMOS, UMR 6158 CNRS, University Clermont Auvergne, 63173 Aubière CEDEX, France.
Sensors (Basel). 2019 Feb 28;19(5):1033. doi: 10.3390/s19051033.
Cyber Physical Systems (CPS) has been a popular research area in the last decade. The dependability of CPS is still a critical issue, and few surveys have been published in this domain. CPS is a dynamic complex system, which involves various multidisciplinary technologies. To avoid human errors and to simplify management, self-management CPS (SCPS) is a wise choice. To achieve dependable self-management, systematic solutions are necessary to verify the design and to guarantee the safety of self-adaptation decisions, as well as to maintain the health of SCPS. This survey first recalls the concepts of dependability, and proposes a generic environment-in-loop processing flow of self-management CPS, and then analyzes the error sources and challenges of self-management through the formal feedback flow. Focusing on reducing the complexity, we first survey the self-adaptive architecture approaches and applied dependability means, then we introduce a hybrid multi-role self-adaptive architecture, and discuss the supporting technologies for dependable self-management at the architecture level. Focus on dependable environment-centered adaptation, we investigate the verification and validation (V&V) methods for making safe self-adaptation decision and the solutions for processing decision dependably. For system-centered adaptation, the comprehensive self-healing methods are summarized. Finally, we analyze the missing pieces of the technology puzzle and the future directions. In this survey, the technical trends for dependable CPS design and maintenance are discussed, an all-in-one solution is proposed to integrate these technologies and build a dependable organic SCPS. To the best of our knowledge, this is the first comprehensive survey on dependable SCPS building and evaluation.
网络物理系统（CPS）是过去十年中热门的研究领域。CPS 的可靠性仍然是一个关键问题，在该领域发表的调查很少。CPS 是一个动态的复杂系统，涉及各种多学科技术。为了避免人为错误和简化管理，自管理 CPS（SCPS）是明智的选择。要实现可靠的自管理，需要系统的解决方案来验证设计并保证自适应决策的安全性，以及维持 SCPS 的健康。本调查首先回顾了可靠性的概念，提出了自管理 CPS 的通用环境循环处理流程，然后通过正式的反馈流程分析了自管理的误差源和挑战。为了降低复杂性，我们首先调查了自适应体系结构方法和应用的可靠性手段，然后介绍了一种混合多角色自适应体系结构，并讨论了在体系结构级别上实现可靠自管理的支持技术。关注以环境为中心的可靠自适应，我们研究了用于做出安全自适应决策的验证和确认（V&V）方法，以及用于可靠处理决策的解决方案。对于以系统为中心的自适应，总结了全面的自我修复方法。最后，我们分析了技术难题中的缺失部分和未来的发展方向。在本调查中，讨论了可靠 CPS 设计和维护的技术趋势，提出了一种集成这些技术并构建可靠有机 SCPS 的整体解决方案。据我们所知，这是关于可靠 SCPS 构建和评估的首次全面调查。
