Rationale-supported mixed-initiative case-based planning

Mixed initiative planning envisions a framework in which automated and human planners interact to jointly construct plans that satisfy specific objectives. In this paper, we report on our work engineering a robust mixed initiative planning system. Human planners rely strongly on past planning experience to generate new plans. ForMAT is a casebased system that supports human planning through the accumulation of userbuilt plans, querydriven browsing of past plans, and several plan functionality analysis primitives. Prodigy/Analogy is an automated AI planner that combines generative and casebased planning. Stored plans are annotated with plan rationale and reuse involves adaptation driven by this rationale. Our system, MICBP, integrates ForMAT and Prodigy/Analogy into a realtime message passing mixed initiative planning system. The main technical approach consists of allowing the user to specify and link objectives that enable the system to capture and reuse plan rationale. We present MICBP and its concrete application to the domain of military force deployment planning. This synergistic system increases the planning efficiency of human planners through automated suggestion of similar past plans and plausible plan modifications.