Full Papers

3.Embodied Intelligent Agents with Cognitive Conscious and Unconscious Reasoning

by Matthew Conforth, Yan Meng (Stevens Institute of Technology)

Abstract: The CIVS (Civilization-Inspired Vying Societies) system is a novel evolutionary learning multi-agent system that uses artificial life methods to produce highly-capable artificial intelligence agents proficient in one or more complex tasks as well as more general adaptability, reasoning, and survivability in dynamic, unpredictable environments. A new cognitive architecture called CHARISMA is being developed as the brain of social agents within the CIVS system. This paper mainly focuses on the development of conscious reasoning and unconscious reasoning mechanisms as well as the interactions between them for the CHARISMA cognitive architecture. Our conscious reasoning design depends on recent advances in Markov logic network algorithms, in particular the online structure learning algorithm. Our unconscious reasoning design is based on Cartesian genetic programming, which we have adapted to work with the semantic hyper network data structure (used for knowledge representation). A novelty-based evolutionary strategy is employed so that the unconscious reasoning is focused on discovering what Markov logic network based conscious reasoning may likely miss. This approach allows interactions between conscious and unconscious reasoning that would not have been possible in the past, and which we believe are vital to achieving more adaptable and creative problem-solving.

Index Terms: cognitive architecture; conscious reasoning; Markov logic networks; unconscious reasoning

Pages: 15-20

Full Text: PDF