The Cognitive Agent Support for Infrastructure Management (CASIM) will be an adaptive and modular architecture that can provide a community of cognitive agents acting in multiple ways to support management of network infrastructure. CASIM agents will incorporate a cognitive module with the ability to synergistically combine multiple modes of cognition. Agents may combine statistical data mining with experience-derived heuristics, or rule-based logic to achieve the level of cognition that is required. Agents may utilize semantic reasoning over abstract concepts and reach conclusions based on matching of a problem to previous experiences. CASIM will incorporate learning & training' mechanisms that allows agents to adapt to the specifics of their local networking environment. Coordination of actions both within and between agents is via a dynamic and hierarchical workflow management system. This, combined with heursitic state estimation logic, will allow CASIM to operate in constrained tactical networks. To achieve the desired degree of autonomy and flexibility is challenging, but not insurmountable. Many of the needed technologies have been developed and demonstrated on earlier projects. By combining the components and providing them with the ontological and logical constructs specific to network management we will be able to achieve the objective capability.
Keywords: Distributed Intelligent Agents, Cognitive Architecture, Machine Learning, Mixed Initiative Reasoning, Workflow Management
