An Infrastructure for Cognitive Architectures

A cognitive agent is a software model of a human's perceptual processes, reasoning, and behavior. A general approach to building cognitive agents must address the problems of knowledge representation, planning and reasoning, sensory and motor modelingwhile satisfying constraints imposed by neural implementations and human cognition. Moreover, different users have varying requirements for their cognitive agents, ranging from modeling fine-grained response and action times to stimuli to modeling highlevel cognitive actions such as reasoning, planning and mission execution. We propose to create a new design and systems engineering methodology, architecture and infrastructure for cognitive agents that promotes reuse of design, behaviors, and code andsupports the incorporation of flexible control strategies and human modeling parameters. It includes both a set of reusable design templates for agents as well as a set of cognitive modules that can be plugged together in different ways to constructdifferent domain-driven cognitive agents. The architecture will be interoperable with existing cognitive approaches through translation and mediation modules as well as through pluggable control and mediation strategies. This cognitive infrastructure andmethodology can be used in conjunction with our distributed software-agent infrastructure, CybeleT. A methodology and cognitive agent infrastructure will allow us to provide our clients with a complete solution to distributed multi-agent simulations.Commercial applications include intelligent agents in computer games and online virtual environments, simulation for measuring human performance operating devices, and modeling human behavior in transportation such modeling pilots and vehicle operators.