CSR: Small: Provably Correct Design of Observation for Fault Diagnosis and State Estimation under Privacy and Network Constraints

Rensselaer Polytechnic Institute, Troy NY

Investigators

Abstract

Information is key in the operation of cyber-physical systems. Consider, for example, the task of designing a monitoring system that can detect faults during the operation of a safety critical cyber-physical system, such as the power grid. To do so, the monitor needs to collect enough information from the cyber-physical system that can distinguish potential faulty operation from normal operation. On the other hand, practicality and privacy limit the amount of information that can be collected. For example, the number of sensors that can be deployed may be limited. Or, the system owner may not wish to share certain information. This project addresses the issue of determining how information can be extracted from the operation of a cyber-physical system to enable fault detection and control, while respecting limitations such as privacy. This work will address a number of fundamental research questions. Q1: Can the right inference be made using the available information, for the purposes of fault diagnosis and state estimation? Q2: What state measurement or observation can be made available to facilitate the answer to Q1 without violating privacy constraints? Q3: When the information is transmitted through a non-ideal communication network, resulting in transmission delay or limited bandwidth, how does it affect Q1 and Q2? This project will develop a framework that provides provably correct answers to all of the questions above. Provable correctness is derived from a model-based approach. The questions are cast in this framework, and obtain provably correct design procedures for state measurement (including software defined sensors and online monitors), state estimation, and fault detection and isolation. The theoretical outcomes from this project will be evaluated and implemented on a number of applications/testbeds, including a smart building testbed, power system networks, and biological signaling networks. Successful outcomes from this project may impact the design and operation of complex cyber-physical systems where distributed sensing, state estimation, and fault detection are used. The evaluation and implementation activities in this project will elucidate more concrete ways the theoretical results can be applied to such systems.