Manuela M. Veloso
This paper introduces our work on mixed-initiative, rationale-supported planning. The work centers on the principled reuse and modification of past plans by exploiting their justification structure. The goal is to record as much as possible of the rationale underlying each planning decision in a mixed-initiative framework where human and machine planners interact. This rationale is used to determine which past plans are relevant to a new situation, to focus user’s modification and replanning on different relevant steps when external circumstances dictate, and to ensure consistency in multi-user distributed scenarios. We build upon our previous work in Prodigy/Analogy, which incorporates algorithms to capture and reuse the rationale of an automated planner during its plan generation. To support a mixed-initiative environment, we have developed user interactive capabilities in the Prodigy planning and learning system. We are also working towards the integration of the rationale-supported plan reuse in Prodigy/Analogy with the plan retrieval and modification tools of ForMAT. Finally, we have focused on the user’s input into the process of plan reuse, in particular when conditional planning is needed.