Advanced Algorithms for Unmanned Systems Resource Optimization

To reduce commander and staff workload in the full spectrum of battlefield management, we propose to create a network of agent modules that enable effective control of multiple robotic elements. In this network, each agent serves as an expert in one or more of the required knowledge areas, and is integrated with other agents, forming a panel of advisors to a central decision-making control agent, the human user, or both. Key requirements for this network include the abilities to understand doctrinal and task requirements; learn new situations and recognize learned patterns; understand commander and enemy intent; collect, filter and fuse relevant information; adapt mixed-initiative interaction; propose optimal solutions when possible and sufficient solutions when necessary; explain reasoning and actions when requested; identify and communicate when user intervention is critical; perform robustly in the face of uncertain, inaccurate or misleading information; and be tolerant of human and system error. Successful advancement toward these requirements will enable an initial simulation-based control system to be developed, that can support future research to refine and develop human-system interfaces and underlying intelligence for embedded autonomy.