attributes. In addition, association learning for diversified attributes from heterogeneous data sources
(sensors and nonsensors) — which may not follow
standard data definitions — can potentially improve
object recognition tremendously (Zhao et al. 2015).
Activity-based intelligence (ABI) and object-based
production (OBP) are computational methods that
offer potential for SA. 2 ABI methods discover new entities and object relationships and behavior patterns
based on the exploitation of all-source data at a massive scale. OBP methods organize and maintain
knowledge around objects to discover things that exist
and things that happen. The combination of these
two methods leads to a tactical decision advantage.
The intended outcome of CID and SA capabilities is
to provide actionable knowledge. Actionable knowl-
edge is knowledge that is accurate, complete, and
timely enough for warfighters to act upon, even to the
point of making COA decisions that include weapon
engagements. Another outcome is to produce knowl-
edge that is shared across distributed warfare plat-
forms, such as ships and aircraft. This shared knowl-
edge enables a battle group to coordinate their actions,
which further improves the tactical advantage.
The observe-orient-decide-act (OODA) loop is the
basis for a conceptual systems of systems (SoS) architecture for a future tactical capability based on the AI
methods in figure 5. The ABI and OBP methods support the ability to observe the real-world AOI. AI
methods and data analytics combine with data
fusion and processing to orient the situation and support decisions and actions. BMAs provide automated
support to complex tactical decisions. The use of networks and multiple instantiations of this architecture
enable shared SA and distributed warfighting coordination.
The desired outcome of the CID and shared SA
processes is actionable knowledge. A notional display
of CID determinations is shown in figure 6. An
important feature would include computed confidence levels to indicate the goodness of the CID
knowledge. This determination would be a combination of complex factors including weather conditions, sensor error, analytical approximation error,
Figure 4. AI Methods for the Knowns and Unknowns.
Known Knowns Known Unknowns