UAV autonomous control system reliability-what should it be, and how do we reach it?

Summary form only given, as follows. The Air Vehicles Directorate of the US Air Force Research Laboratory is pressing ahead with designing and implementing autonomous control systems for both individual UAVs as well as hierarchical and/or distributed control for packs (or swarms) of UAVs. The goal is to truly make the man the stategist, tell the UAV(s) what is needed and the UAVs go away, figure out how- to do it, and execute the mission. Implicit in all of this is safety. One of our goals is to make these UAVs as reliable as, or better than, manned aircraft. This paper looks at what that actually means without a pilot in the loop, and the requirements flow-down, both overt and implied, which places the specifications on the subsystems and components. At the end we "prognosticate" the possible impacts to current best practices in testing and certification, stressing the need to possibly look at this a bit differently for these types of systems-a paradigm shift may be in order.