An Ensemble-Based Approach for Scalable QoS in Highly Dynamic CPS

Abstract

Modern cyber-physical systems (CPS) often involve distributed devices/components that closely interact with each other and their environment. In this context, operation conditions may constantly change and it is not always possible to guarantee quality of service (QoS), particularly, if resourcesdegrade or stop being available. In addition, sometimes, one would like QoS to scale up/down with operation conditions, e.g., maximize efficiency, minimize energy consumption, etc. without compromising safety. However, traditional design and development techniques fail to capture the dynamics of modern CPS, since they rather focus on individual components/devices, and are unable to provide such QoS guarantees. To overcome this problem, we propose a design methodology based on the concept of ensemble, i.e., a dynamic grouping of components, which allows for scalable QoS guaranties. We illustrate the utility of our approach based on a case study consisting of an intelligent production line and analyze the effect on performance as communication between components degrades. Finally, our methodology can be incorporated into existing ensemble-based tools such as DEECo, Helena or jRESP to generate executable code to be deployed onto distributed devices.