Artificial Intelligence Techniques for Reactive Planning

We propose an innovative combination of artificial intelligence (AI) techniques to meet NASA's scheduling needs for representation, schedule design and schedule analysis. We will use AI knowledge representations including frames, semantic networks and rules for capturing schedule components and constraints. We will facilitate schedule design and reactive planning through the use of the intelligent entities concept and similarity-based and constraint satisfaction techniques, techniques we have used with much success for NASA's high-level manifest planning. We will enhance NASA's schedule analysis capabilities through the application of AI constraint-based techniques, capture and implementation of heuristic analysis techniques and standard graph negotiation techniques. NASA will benefit from improved schedule maintenance, display and analysis capabilities. The goals of the Phase I research are to thoroughly define the planning domain, elaborate the artificial intelligence techniques useful for automating the scheduling problem, prove the feasibility of these techniques, and develop a complete system specification for an automated NASA scheduler. Phase I research and prototype development will lay the groundwork for the Phase II implementation of an automated scheduling and analysis system and the eventual commercialization of a reactive planning system.