A PLANNER FOR THE CONTROL OF PROBLEM-SOLVING SYSTEMS

As part of research on sophisticated control for sensor interpretation , we have developed a planning-based control scheme for blackboard sys tems. A planner's goal/plan/subgoal structure provides explicit contes t information that can be used to index and apply large amounts of con text-specific control knowledge. The key obstacle to using planning fo r the control of problem solvers is the need to deal with uncertain an d dynamically changing situations without incurring unacceptable overh ead. These problems have been addressed in several ways: our planner i s script-based, planning and execution are interleaved, plans can invo ke information gathering actions, plan refinement is controlled by pla n-specific focusing heuristics and the system's focus of attention can be dynamically shifted by the refocusing mechanism. Refocusing makes it possible to postpone focusing decisions and maintain the opportunis tic control capabilities of conventional blackboard systems. Planning with refocusing results in a view of the control process as both a sea rch for problem solutions and a search for the best methods to determi ne these solutions. The planner has been implemented in the RESUN (REs olving Sources of UNcertainty) interpretation system and has been used with a simulated aircraft monitoring application and a system for und erstanding household sounds. Our experience confirms that the combinat ion of control plans with context-specific focusing heuristics provide s a modular framework for developing and maintaining complex control s trategies. In experiments, we have been able to achieve significant pe rformance improvements as a result of the ability to encode sophistica ted control strategies despite the overhead of the planning mechanism.