Architecture for Predictable Mixed-Initiative Planning

We propose to develop an innovative software architecture for real-time planning, monitoring, and control that includes as a core component a rich capacity for human interaction and collaboration. Because developing such a system represents a formidable challenge, we suggest an integrated approach drawing upon a broad range of AI techniques. Advanced hierarchical planning and scheduling algorithms will permit the system to quickly create complex plans satisfying intricate time and resource constraints. Automatic truth maintenance technology will furnish the ability to clearly explain the reasoning behind its actions. The proposed architecture will also incorporate a case-based reasoning engine allowing it to demonstrate predictable behavior by choosing current actions based on past actions in similar situations. Finally, the system will employ an adaptive intelligent interface that will provide unprecedented eas-of-use, allowing non-technical users to supply advice and guidance without a great deal of special training. The resulting system will enhance crew performance by furnishing sophisticated automated planning, monitoring, and control capabilities, while leaving ultimate authority in human hands. We will absolutely demonstrate the feasibility of our ideas through the development of a Phase I, proof-of-concept prototype.