CPS: Medium: The Ectokernel Approach: A Composition Paradigm for Building Evolvable Safety-critical Systems from Unsafe Components
Lead PI:
Tarek Abdelzaher
Co-PI:
Abstract
The objective of this research is to develop a new approach for composition of safety-critical cyber-physical systems from a small code base of verified components and a large code base of unverified commercial off-the-shelf components. The approach is novel in that it does not require generating the entire code base from formal languages, specifications, or models and does not require verification to be applied to all code. Instead, an explicit goal is to accommodate large amounts of legacy code that is typically too complex to verify. The project introduces a set of verified component wrappers around existing unverified code, such that specified global system properties hold. The intellectual merit of the project lies in its innovative approach for managing component interactions. Unexpected interactions are the primary source of failure in cyber-physical systems. A fundamental conceptual challenge is to understand the different interaction spaces of cyber-physical system components and determine a set of trigger conditions when certain interactions must be restricted to prevent failure. The project develops analysis techniques that help understand the different interaction types and provides component wrappers to restrict them when necessary. Broader impact lies in significantly reducing the design and composition effort for the next generation of safety-critical embedded systems. A variety of student projects are being offered to undergraduates and graduate students. The researchers especially encourage women and minorities to participate. Outreach activity, such as hosting K-12 students on school field/science days, further strengthen the educational component. Technology transfer to John Deere is expected.
Performance Period: 08/15/2010 - 07/31/2014
Institution: University of Illinois at Urbana-Champaign
Sponsor: National Science Foundation
Award Number: 1035736