openCPS: Stochastic Models for the Design and Analysis of Open-Ended Cyber-Physical Systems

点击翻译按钮获取中文摘要

Cyber-physical systems (CPS) typically consist of one or more embedded devices that implement feedback loops to control processes in the physical world; thereby physical processes influence computations and vice versa. In contrast to traditional embedded control systems, CPS are primarily concerned with exploiting the interplay between physical dynamics and computation towards improving safety, quality, and efficiency among others.There is a wide range of applications that can be classified as CPS. Hence, specialized design and development approaches have originated. The great majority of these approaches focus on closed or bounded systems, i.e., those where the system's parts or components are fixed and known a priori. However, there are a number of recently upcoming applications such as smart grid, intelligent transportation systems, smart factories, etc., in which systems are often open-ended. That is the case where not all system's parts or components are known prior to runtime and, as a consequence, the system cannot be fully described during design.Open-ended CPS usually rely on autonomous mobile devices that form an ecosystem. The joint operation within this ecosystem provides functionality which is otherwise unattainable by single devices in isolation. As a result, there is a strong need for design and development techniques and methods that focus on the ecosystem as a whole. On the one hand, these techniques and methods need to provide systematic software engineering practices that allow managing the high complexity and help controlling so-called emergent behavior. On the other hand, they should support and ease the analysis and validation of timing and reliability properties, in particular, in the context of safety-critical applications where certification is required.Traditional design and development techniques from the embedded domain are normally of deterministic nature and, hence, less adequate for open-ended CPS which imply certain randomness, i.e., components may join and leave the system at arbitrary points in time. In contrast, the use of stochastic models allowing statistical and probabilistic techniques for timing/reliability analysis and validation seems to be more meaningful. Lately, in the literature, such methods have been proposed for estimating the WCET (worst-case execution time) of tasks and providing probabilistic schedulability guarantees in the presence of errors. However, these works focus on stochastically modeling a small and relatively isolated part of the system, viz., computation on one processor. In this project, we are concerned with the use of stochastic models for system-level design and analysis of open-ended CPS. We intend to investigate on approaches to incorporate randomness into existing computation and communication paradigms making them more amenable to stochastic modeling and, hence, to statistical and probabilistic design and analysis methods.