Intelligent Automation, Inc.

Research and Development » (old) Distributed Intelligent Systems » Expertise » Multi-agent Systems and Architectures » 

I n this effort we are developing a new framework and methodology, called MaGIC, for creating cognitive systems and architectures and to develop a set of software tools and methodology to support the approach. The approach seeks to address the challenging systems development problem faced by designers of large-scale cognitive systems. We focus particularly on the interface issues between knowledge representations, execution models and control structures that will enable interoperability, integration and design of hybrid architectures. The MaGIC component model provides a way to encapsulate and reuse cognitive functions and processes and to express the interactions between autonomously executing components. It also provides a conceptual framework that enables the creation of multiple types of cognitive architectures. Representation scheme currently included in MaGIC are - ontology, schema specification, constraints and rules, language bindings and wire protocol.

As a part of this effort we are also developing a software toolset and runtime environment that supports the MaGIC component-based approach using an agent-based simulation infrastructure. The infrastructure is intended to provide a predictable discrete time simulation platform and supported distributed and parallel CPU configurations. The framework and tools will provide users the following capabilities:

• Iterative design of cognitive systems through composition
• Integration of hybrid control structures and knowledge representations
• Protocol design and validation/verification tools
• A test and simulation environment that enables discrete time repeatable simulation.
• A general methodology for the design and implementation of component-based hybrid cognitive systems.

This effort is supported, in part, under contracts from DARPA