Safety-Feature Modeling and Adaptive Resource Management for Mixed-Criticality Cyber-Physical Systems

pdf

Abstract:

This project is concerned with ensuring operational safety of complex cyber-physical systems such as automobiles, aircraft, and medical devices. Modern development techniques for such systems rely on independent implementation of safety features in software and subsequent integration of these  features within system platform architectures. The current trend in developing these systems, driven by the need to reduce cost and energy consumption, is to share computational resources between  different features. The goal of this project is to develop techniques to predict possible interactions between features, detect them in the features' concrete implementations, and either eliminate or mitigate these interactions through precise modeling and enforcement of mixed-criticality cyber-physical system semantics. While the project is developing general purpose techniques applicable to different application domains, work concentrates on automotive systems as case studies. An industrial collaborator (not supported by NSF funds) is providing domain expertise to ensure practical applicability of results. The project currently pursues two related research directions:

  1. Criticality-sensitive deployment of features on distributed platforms. A car can be exploited in different scenarios, in which the same features are used, but their roles and, therefore, criticality levels are different. By treating scenarios as system modes, we study means of allocating runnables from different features to processors on the platform in a way that would enable criticality levels to be respected in all system modes.
  2. Platform Support for Real-Time Retargetable Virtualization.  Our approach leverages the recent development of RT-Xen, a real-time patch for the popular Xen virtualization platform. Timing isolation provided by RT-Xen-like platforms enables us to support middleware-aware applications, such as AUTOSAR components in the automotive domain, by ensuring location transparency and end-to-end real-time guarantees.
  • adaptive scheduling
  • automotive safety features
  • General Motors
  • multi-criticality
  • University of Pennsylvania
  • Washington University in St. Louis
  • Architectures
  • Architectures
  • Automotive
  • CPS Domains
  • Platforms
  • Modeling
  • Transportation
  • CPS Technologies
  • Foundations
  • National CPS PI Meeting 2015
  • 2015
  • Abstract
  • Poster
  • Academia
  • Industry
  • 2015 CPS PI MTG Videos, Posters, and Abstracts
Submitted by Oleg Sokolsky on Sat, 01/30/2016 - 23:45