TRANSFORMATIONAL SYNTHESIS AND REASONING ABOUT THE FUTURE

TWO KEY PROBLEMS FOR BUILDING PRACTICAL YET GENERAL KNOWLEDGE-BASED PLANNING APPLICATIONS ARE 1) REPRESENTING THE PLAN AND 2) PROJECTING THE PROBABLE EFFECTS OF EXECUTING THE PLAN. BOTH OF THESE PROBLEMS REQUIRE TECHNIQUES FOR REASONING ABOUT THE FUTURE. THE REPRESENTATION AND THE PROJECTION PROBLEMS BOTH NEED TO BE SOLVED FOR THE PARTIAL AND ABSTRACT PLANS THAT WILL BE GENERATED AT INTERMEDIATE STAGES OF THE PLANNING PROCESSES AS GOALS AND CONSTRAINTS ARE EVOLVED TOWARD AN EXECUTIBLE PLAN. WE PROPOSE TO EXTEND OUR CURRENT TRANSFORMATIONAL SYNTHESIS METHODOLOGY TO INCLUDE GENERAL SOLUTIONS FOR THESE PROBLEMS.TRANSFORMATIONAL SYNTHESIS IS AN INSTANCE OF WHAT WE BELIEVE IS AN EMERGING CONSENSUS METHODOLOGY FOR INTEGRATING MULTIPLE PLANNING TECHNIQUES IN A WAY THAT ALLOWS THE MOST APPROPRIATE TECHNIQUES TO BE USED FOR EACH ASPECT OF THE APPLICATION PROBLEM. IN PARTICULAR, WE PROPOSE TO EXTEND THE PLAN REPRESENTATIONS USED IN PAST APPLICATIONS OF TRANSFORMATIONAL SYNTHESIS TO SUPPORT CONSTRAINT-BASED REASONING ABOUT DEPENDENCIES BETWEEN PLAN STEPS (INCLUDING TEMPORAL DEPENDENCIES) AND TO BE ABLE TO PROJECT THE EFFECTS OF PARTIAL AND ABSTRACT PLANS FOR THE CASE WHERE THE STATE SPACE REPRESENTATION HAS COMPLEX INTERNAL DEPENDENCIES. PROJECTION WILL HANDLE UNCERTAINTY IN KNOWLEDGE ABOUT THE WORLD SITUATION AND ABOUT THE EFFECTS OF THE PRIMITIVE OPERATIONS INVOLVED IN THE PLAN.