Perception Subsystem for a Cognitive Architecture

A new computational perception algorithm is introduced for intended integration into a new cognitive architecture. The subsystem extracts features describing relations that hold over two or more entities in a simulated microworld which it observes. A new feature extraction algorithm used in the subsystem, the Adaptive Parallel Terraced Scan, is described in detail. Initial experiments which begin to investigate performance of the proof-of-concept prototype are described, and results presented. Finally, areas of future work to be pursued in order to build on these initially promising results are outlined.