Towards Provably Correct Design of Human-automation Systems: Hybrid System Observability and Reachability
Speaker: Meeko Oishi, University of New Mexico
Department: Mechanical & Aerospace Engineering
Location: Bowen Hall Auditorium 222
Date/Time: Friday, May 3, 2013, 3:30 p.m. - 4:30 p.m.
In many complex cyber-physical systems, human interaction with coupled cyber and physical components can significantly complicate system safety. Such systems are often large enough that simple intuition is not enough to determine whether the user-interface, a device that both provides information to the user about the underlying automation and allows the user to issue input commands to the system, as well as the corresponding automation, is correctly designed. Consider, for example, automation surprises and other mode errors that can occur in flight management systems, despite extensive simulation and experimental testing. We propose the development of observability and reachability techniques to create a new level of confidence and reliability in safety-critical cyber-physical systems, by predicting, at the design stage, configurations under which failures might occur. Observability techniques can determine whether the user has adequate information to accomplish a known task; reachability techniques can prevent the system from reaching configurations known a priori to be unsafe. Such control theoretic techniques could form the basis of design aids for provably correct human-automation systems.