Research

Adaptable Autonomous Systems (AAS) are autonomous systems capable of adaptation with respect to changes in their state and/or their environment. The dynamic and unfamiliar nature of the environment, let alone safety-critical necessitates employing runtime safety assurance methodologies. Besides, random failures or malfunctions in system components transition the system to an unsafe state which further raises concerns for its safe behavior. The intensity of hazard(s) caused by these malfunctions depends upon the current state of the operational context in which they occur. Thus to guarantee safe behavior of the system, one must be aware of its operational context in the first place.

Traditional safety assurance demands complete system information to be available at design time; from operational behavior of the system to the environment in which it is functioning. Since AAS adapt their behavior in response to changes in their own state and/or their operational environment, this information cannot be fully determined well in advance. As a result, traditional safety methodologies, despite still being vital, are inadequate to ensure safe behavior. This bring forth two essential facets to consider for safety assurance of AAS: First, shifting conventional design time safety approaches to runtime safety methodologies and second, an explicit incorporation of system-context association in these runtime methodologies.

My research involves understanding and modeling the operational context of an AAS from a safety perspective. I try to find out the system-context association critical for safety assurance for AAS. It aids in answering crucial questions like: Why system must be aware of its context? What must be done with the collected knowledge about the context? How and where this awareness should be implemented?