Runtime Safety Evaluation in Autonomous Systems (ReSonAte) is a framework for estimating the dynamic risk of Autonomous Cyber-Physical Systems. It uses the design-time hazard analysis information to build btd which describe potential hazards to the system and how common events may escalate to consequences due to those hazards. The risk posed by these hazards can change dynamically since the frequency of events and effectiveness of hazard controls may change based on the state of the system and environment. To account for these dynamic events at runtime, ReSonAte uses design-time Bow-Tie Diagram models along with information about the system's current state derived from system monitors (e.g. anomaly detectors, assurance monitors, etc.) and the operating environment (e.g. weather, traffic, etc.) to estimate dynamic hazard rates. The estimated hazard rate can be used for high-level decision making tasks at runtime, to support self-adaptation of Cyber-Physical Systems.
This work is supported in part by the DARPA Assured Autonomy program.
ORGANIZATION
Vanderbilt University, Nashville, TN, USA
Charles Hartsell
Shreyas Ramakrishna