Recently, several new and efficient approaches have been developed for plan synthesis that relate plan synthesis to constraint satisfaction. At present, there exist a huge gulf between these new breed of algorithms and the traditional refinement planning approaches. The PI has developed a generalization of the refinement planning framework within which the newer approaches mentioned can all be seen as particular instances of a distinguished subclass of refinement planners called "disjunctive planners." This insight opens up several exciting avenues of focused research: Could the missing planners corresponding to the other instances of general disjunctive planning be more efficient? What are the tradeoffs governing the efficiency of these spectrum of planners? How does one effectively transfer the progress made in improving the performance of non-disjunctive planners to disjunctive planners? The PI proposes several specific research tasks to answer these questions. The proposed disjunctive planning framework bridges the areas of constraint satisfaction, planning and propositional reasoning, and promises to offer a radically efficient substrate for plan-synthesis. The framework bridges the areas of constraint satisfaction, planning and propositional reasoning, and promises to offer a radically efficient substrate for plan-synthesis. The framework also supports seamless integration of planning and scheduling.
