CAREER: Robustness Guided Testing and Verification for Cyber-Physical Systems
Lead PI:
Georgios Fainekos
This project develops a theoretical framework as well as software tools to support testing and verification of a Cyber-Physical System (CPS) within a Model-Based Design (MBD) process. The theoretical bases of the framework are stochastic optimization methods, and robustness notions of formal specification languages. The project's research comprises three components: development of conditions on the algorithms and on the structure of the CPS for inferring finite-time guarantees on the randomized testing process; the study of testing methods that can support modular and compositional system design; and investigation of appropriate notions of conformance between two system models and between a model and its implementation on a computational platform. All of these components are needed to support testing and verification in all the stages of an MBD process as well as to support component reuse, incremental system improvements and modular design. The evaluation of the framework is driven by the problems of verifying automotive control systems and medical devices. As safety-critical CPS become ubiquitous, the need for design methods that guarantee correct system functionality and performance becomes more urgent. Certification and government agencies need dependable testing and verification tools to incorporate in certification standards and procedures. The concrete benefits to the society are both in terms of reduced catastrophic design errors in new products and in terms of reduced economic costs for new product development. The former increases the confidence in new technologies while the latter improves the competitiveness of the companies that utilize such technologies. The theoretical results of this project are being incorporated into software tools for testing, verification and validation of complex CPS. The evaluation focus of the project on verifying infusion pumps and automotive control software ultimately helps in avoiding harmful losses due to errors in these safety-critical systems. The use of any software tool that is based on formal or semi-formal methods requires engineers with solid training on these technologies. This proposal puts forward an education curriculum for developing new courses that introduce formal and semi-formal methods for CPS at all levels of higher education, i.e., undergraduate, graduate and continuing education. Particular attention is devoted into on-line continuing education of practicing engineers who must acquire new MBD skills.
Performance Period: 08/15/2014 - 07/31/2019
Sponsor: National Science Foundation
Award Number: 1350420