a new framework for cognitive information fusion
Key Points
- DAIRNet Theme: Autonomous Processing and Reasoning
- Griffith University
- Point of contact: Prof Ted Goranson
Our approach builds general logical forms for ‘action statements’ built on what the end user will need. An end user here is a mission specialist or weapon/EW/Cyber operator. These are semantic-free structures with causal types, designed to host representations of what a user would need to know without specifying any details. They can accommodate unknown and unexpected situations and comprehending deception. These abstract mathematical structures are designed to be accessed by aggregators early in the food chain. The aggregation options are greatly constrained, making this feasible. Some of these aggregators can be autonomous.
This is a top-down organization method with several benefits: it integrates with ML layers, building adaptive layers to make them responsive to unexpected change; it induces logical, causal, ‘what-if’ and ‘why’ statements; because the type system is abstract, it bridges finegrained parametric data and late-stage semantic knowledge, and, because it can retains the geospatial nature of much of the finegrained sensor data, the reports can be visualized using spatial metaphors if desired. It works with existing ML systems with little disruption.
