High Assurance for Distributed Cyber Physical Systems

Scott Hissam, Sagar Chaki and Gabriel Moreno, Proceedings of the Architecting Self-Managing Distributed Systems Workshop (ASDS'15), Sep 7, 2015.

Abstract:Distributed Adaptive Real-Time (DART) systems are interconnected and collaborating systems that continuously must satisfy guaranteed and highly critical requirements (e.g., collision avoidance), while at the same time adapt, smartly, to achieve besteffort and low-critical application requirements (e.g., protection coverage) when operating in dynamic and uncertain environments. This short paper introduces our architecture and approach to engineering a DART system so that we achieve high assurance in its runtime behavior against a set of formally specified requirements. It describes our technique to: (i) ensure asymmetric timing protection between high- and low-critical threads on each node in the DART system, and (ii) verify that the self-adaptation within, and coordination between, the nodes and their interaction with the physical environment do not violate high and low criticality requirements. We present our ongoing research to integrate advances in model-based engineering with compositional analysis techniques to formally verify safetycritical properties demanded in safety-conscious domains such as aviation and automotive; and introduce our DART model problem, which serves to demonstrate of our engineering approach.