An Agent-Based Component Architecture for Cognitive Decision Support

In this SBIR Phase I effort, we propose an architectural framework and design methodology for cognitive decision aids based on software agents which addresses the fundamental challenge of creating agile decision support systems that can be rapidly adapted for evolving system configurations and requirements. The basis of this approach is an agent-based component model and a library of formally defined protocols and roles that describe the interactions between cognitive decision components. During system execution, each component is instantiated as a software agent with specified execution properties. The system components include bounded-time reasoning and planning components, sensor fusion components, and human interface components. The approach will support flexible composition of decision aid systems using reusable cognitive components. We will further demonstrate how an integrated software development tool-chain can be used to enable rapid design, testing and realization of different system configurations. The approach will be illustrated using in a prototype component model, architecture specification and decision-aid use case using in a networked fires application. In this Phase I effort, we will focus on formulation the overall methodology and architecture for creating agile, adaptable decision aid components and systems.