Toward Dependable Cyber-Physical Systems : a Study on Design Pattern and Evaluation Methodology

Abstract

People live in and interact with the physical world. As technology advances, embedded devices with sensing, computation, actuation, and networking capabilities are drastically changing our way to interact with the physical world. This introduces a new type of systems, namely, Cyber-Physical Systems (CPS). CPS tightly integrates discrete computing and continuous-time physical-world entities. It reshapes the way that humans interact with the physical world, and thus is believed to have deep social and economical impacts. As many CPS applications are mission/life critical, dependability is a top concern. To build dependable CPS, various fault prevention, fault tolerance, and fault removal measures are needed in the new context of CPS. In this thesis, we address several challenging issues on building dependable CPS. First, we propose a fault prevention solution to guarantee Proper-Temporal-Embedding (PTE) safety rules in wireless CPS. The proposed solution exploits the leasing design philosophy to tolerate arbitrary wireless communication failures, and support real-time temporal constraints. The proposed solution is validated by two case studies: one on medical CPS and the other on control CPS. The performance of our solution is also compared to a polling based solution. Simulation results show that our proposed solution achieves better user experience when wireless channel is benign or moderately adverse, and better resource usage in all scenarios. Second, we propose a cross-domain noise profiling framework for control CPS. The proposed framework plays a key role in control CPS dependability evaluation, an essential tool to CPS fault tolerance and fault removal. Key elements of this framework include a hybrid automata reachability based dependability metric, and a Lyapunov stability theory based benchmark shrinking strategy. Case studies are carried out to validate the proposed framework and showcase its usage.

Date of Award	2016
Original language	English
Awarding Institution	The Hong Kong Polytechnic University
Supervisor	Qixin Wang (Chief supervisor) & Lei Zhang (Co-supervisor)