Journal de la gestion de la défense
Libre accès

Abstrait

Sparse Requirements Systems Engineering and Implications for Assured Autonomy

Gregory Gosian, Kelly Cohen

Systems engineering in program execution identify comprehensive requirements definition as a prerequisite for successful programmatic execution. Conversely, not fully defining system requirements is frequently branded as being the root cause for programmatic failures. The failures manifest themselves as not meeting performance thresholds, exceeding budget, or delivering late. With respect to executing engineering programs in the 21^st century, and especially for the Department of Defense (DoD) and other Government customers, there are drivers that require systems engineers to commence system specification prior to fully defining (or having defined for them) all of the system or mission requirements. As such, there is a need for a new method of systems engineering that allows for progress in environments of uncertainty – a Sparse Requirements Systems Engineering (SRSE) paradigm. Further, a consequence of architecting Learning Enabled – Cyber Physical Systems (LE-CPS) in such a reduced requirements environment is the need to assure safe, intended operation. SRSE emulates an incomplete training data set of inputs, further increasing the potential for unintended response and operations of such LE-CPS.