Dynamic Assurance Cases for Learning-enabled Systems
The pace of innovation in machine learning technologies and their increasing use in learning-enable components (LECs) in safety and mission-critical applications, e.g., deep neural networks used for perception in self-driving road vehicles, currently far outstrips that of the applicable regulatory and standardization efforts to create the bases against which it would be established that the resulting systems can be relied upon. Risk-based approaches to engendering trust, in the form of argument-based safety cases, have shown promise for the assurance and subsequent operational approval of novel systems.
However, LECs pose particular challenges for certification, as does the gap between the state of the art in safety assurance, and how aviation systems are certified in practice. Towards straddling this gap, we are developing the dynamic assurance case (DAC) concept as a model-based, multifaceted approach to the assurance of LEC-based systems. Our vision is one of a rich, expressive, and formally-founded framework, going well beyond how argument-based safety cases are currently developed. In particular, besides recording assurance rationale in a modular fashion, DACs: i) capture assurance policies and a conforming assurance architecture, ii) provide a framework for assurance quantification, and iii) also supply the means to admit design-time verification and validation (V&V) evidence, along with run-time evidence from operational monitoring.
Time: Monday June 17, 14:30-15:30, followed by coffee and cookies.
Place: Lecture room EL41, EDIT building, Maskingränd 2
Welcome! (no registration required)
Dr. Ewen Denney is a senior computer scientist and the Technical Area Liaison for the Robust Software Engineering Group of the Intelligent Systems Division at the NASA Ames Research Center in Mountain View, California. He currently leads a research group that is seeking to establish a rigorous basis for safety assurance cases, develop tool support, and apply this work to NASA problems.
The current focus of his work at NASA is transitioning assurance case technology to the US aerospace industry. He also is PI on a DARPA-funded project that aims to extend and apply these techniques to autonomous systems based on AI. He is the author of numerous publications on formal methods, program synthesis, and safety assurance cases, and has chaired several international conferences. In particular, he was the founding co-chair of the NASA Formal Methods Symposium, in 2009, and founded the International Workshop on Assurance Cases for Software-intensive Systems, which he has co-chaired each year since 2013.