eduzhai > Applied Sciences > Engineering >

Quantifying Assurance in Learning-enabled Systems

  • Save

... pages left unread,continue reading

Document pages: 16 pages

Abstract: Dependability assurance of systems embedding machine learning(ML)components---so called learning-enabled systems (LESs)---is a key step fortheir use in safety-critical applications. In emerging standardization andguidance efforts, there is a growing consensus in the value of using assurancecases for that purpose. This paper develops a quantitative notion of assurancethat an LES is dependable, as a core component of its assurance case, alsoextending our prior work that applied to ML components. Specifically, wecharacterize LES assurance in the form of assurance measures: a probabilisticquantification of confidence that an LES possesses system-level propertiesassociated with functional capabilities and dependability attributes. Weillustrate the utility of assurance measures by application to a real worldautonomous aviation system, also describing their role both in i) guidinghigh-level, runtime risk mitigation decisions and ii) as a core component ofthe associated dynamic assurance case.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×