| This research work is focused on modeling and design of various memory systems and processes in cognitive software agents. Human memory seems to come in myriad forms: sensory, procedural, working, declarative, episodic, semantic, long-term memory, long-term working memory and possibly others as well. This research attempts to model transient episodic memory, declarative memory and the consolidation process from transient episodic memory to declarative memory. This modeling is done in the context of cognitive software agents like IDA, the Intelligent Distribution Agent built for the U.S. Navy. The goal of this research is to develop some of the memory tools needed to build intelligent software systems, while attempting to understand how the human cognitive processes come about.; We have designed and implemented a modification to Kanerva's Sparse Distributed Memory (SDM) architecture. The modification provides significant performance improvement over the original SDM. This architecture promises to be an ideal candidate for memory systems with partial writes as well as partial read-cues. We have also implemented several decay mechanisms for this modified architecture, to enable its use as Transient Episodic Memory (TEM) in 'conscious' software agents. Further, we have proposed a design for memory-consolidation from TEM to declarative memory (DM) in cognitive software agents. |
