Cognitive Agent Architecture for Multi-System Planning (CAAMP)

We propose to develop a Cognitive Agents Architecture for Multi-system Planning (CAAMP) that will automate Navy multi-system planning. CAAMP comprises an agent collaboration framework, planning agents, and tactical decision aids. Each planning agent will be generating (sub)plans in response to requests from human planners or other planning agents. Collaboration is necessary because complex missions, in general, cannot be planned in a strictly top-down, sequential manner.This multi-agent approach has several advantages. First, it is easier to develop, understand, and test smaller, simpler automated planners that each solve a portion of the planning problem, compared to a single, complex system. Second, each planning agent can use the most appropriate knowledge representation and reasoning method for the planning subproblems it is designed to solve. Third, each agent can be aligned with a human planning cell to streamline knowledge engineering, user testing and validation, user-system collaboration, and knowledge maintenance. Finally, a modular framework that enables loosely coupled agents to be updated or replaced by different contractors can leverage ongoing planning tool investments more effectively and can simplify insertions of new technologies. During Phase I, we will design, prototype, and demonstrate the feasibility of developing a capable strike mission planner composed of collaborating agents.