Patterns for Self-Adaptation in Cyber-Physical Systems

Abstract

Engineering Cyber-Physical Systems (CPS) is challenging, as these systems have to handle uncertainty and change during operation. A typical approach to deal with uncertainty is enhancing the system with self-adaptation capabilities. However, realizing self-adaptation in CPS, and consequently also in Cyber-Physical Production Systems (CPPS) as a member of the CPS family, is particularly challenging due to the specific characteristics of these systems, including the seamless integration of computational and physical components, the inherent heterogeneity and large-scale of such systems, and their open-endedness. In this chapter we survey CPS studies that apply the promising design strategy of combining different self-adaptation mechanisms across the technology stack of the system. Based on the survey results, we derive recurring adaptation patterns that structure and consolidate design knowledge. The patterns offer problem-solution pairs to engineers for the design of future CPS and CPPS with self-adaptation capabilities. Finally, the chapter outlines the potential of Collective Intelligence Systems for CPPS and their engineering based on the survey results.

An Overview of Design Patterns for Self-Adaptive Systems in the Context of the Internet of Things

“…However, design patterns for multi-agent systems are applied in several applications. Musil et al [30] propose new design patterns capturing best practices for self-adaptation in CPSs. Giese et al [31] describe several architectural patterns that describe reflection in self-aware computing systems.…”

“…Content-based Routing aims at routing messages within a distributed monitoring setting according to the message's content. Ramirez specifies the design patterns by using structural UML diagrams and sequence diagrams.Musil et al propose three design patterns for selfadaptation, focusing on CPSs: Synthesize-Utilize, Synthesize-Command, and Collect-Organize[30]. Precisely, these multiadaptation patterns capture information on the used adaptation mechanism, the interactions between those mechanisms across layers, and a definition of layers (i.e., physical, proxy, communication, service middleware, application, and social layer).…”

“…Two of them, i.e., Coordinated Control and Information Sharing, provide a decentralized approach, while the other three are hybrid. The patterns proposed by Musil et al[30] aim to improve the utility of the services of distributed applications using MAPE-K control loops by exploiting context information.While the patterns proposed by Weyns et al and Musil et al have a global view on the self-adaptive mechanisms (i.e., MAPE-K control loop level), Ramirez et al…”

Industrial Use Cases of Cyber Physical Systems in EU Projects: Preliminary Study

“…The current surveys and systematic reviews mostly target CPS architectures (i.e., decomposition into smaller parts and their interconnection) and their applications, which is a well-known method to deal with the increasing complexity of those systems. This includes various aspects, such as Architectural Description Languages (ADL) [16] (called architectural languages (AL) in the paper) and self-adaptation [7], as well as many others [15] [12] [11]. For instance, in [17], the authors are interested in identifying the application domains, challenges, goals, and solutions of architecting CPSs.…”

“…The systematic study found links between challenges and the suggested solutions for architecting CPSs, which provided a clear idea as to what kind of research could be targeted in the future. In contrast to this, the researchers in [7] headed for finding out application domains and uncertainty types in self-adaptive CPS, and they were able to synthetize patterns from multiple adaptation mechanisms and identify on which layers of the CPS architecture they are applied. As a result of this study, the patterns present an engineering method to be used in designing future self-adaptive CPS.…”