Mixed Initiative Planning and Control Under Uncertainty

A hierarchical architecture is developed to provide closed-loop, mixed initiative planning and control for distributed force teams of unmanned air vehicles in an uncertain military operational environment. The hierarchical architecture derives from a rigorous decomposition of the problem that preserves systemlevel objectives while respecting local constraints and defines the interactions and information exchanges between decision-making nodes at each level. An intelligent adversary is addressed in planning and decision-making through coupling of uncertainty in state estimation and the risk associated with possible system states. The proposed hierarchical structure also accommodates human decision-makers and operators at any level within any planning and control function. This is made possible by the incorporation of humancentered design principles and human behavior representation models that enable human operators and machine automation to function as a cooperative team. Game-theoretic estimation and control techniques capture the actions of an intelligent adversary in order to improve performance under imperfect state knowledge. The problem decomposition and the use of experimentally derived heuristics make this approach computationally tractable. Computational cognitive process models capture expert human decision-making, thereby providing a foundation for bridging the gap between engineering estimation/optimization algorithms and naturalistic human machine interfaces (HSIs) that support effective mixed initiative monitoring, planning, and control in dynamic environments.