MIXPLAN: A CLP-Based Mixed-initiative Planning System for Temporal Domains

In mixed-initiative planning the objective is to achieve the best combination of human and machine involvement in the process of constructing qualitative plans. Several mixed-initiative planning systems have been developed in recent years, aiming at reinforcing planning operators in constructing plans for sophisticated domains, such as military or space. In this paper we present a lightweight mixed-initiative planning system, called MIXPLAN, which targets an entirely different group of people, i.e. students studying temporal partial-order plan-space planning. MIXPLAN helps a student to solve temporal planning problems, by presenting her continuously with the available options at each choice point, propagating the results of her choices, automatically selecting on behalf of the user when a single viable choice exists, allowing the user to backtrack to any choice point and automatically backjumping in case of failure. All these features are coupled with a graphical representation of the partial-ordered plan. The system has been implemented on top of ECLiPSe Constraint Logic Programming system platform, employing Graphviz for visualization purposes.