Developing a General Contingency Planner for Adversarial Planning
Abstract:
This report summarizes the first year of effort to investigate the basic mechanisms for solving planning problems in environments that contain intelligent adversaries. A general purpose planner that can solve adversarial planning problems in a variety of domains is presently under development. This planner, called CPx Contingency planner version x, uses a formalism for representing the plans, which are consistent with the goal tree formalisms found in action planning in Artificial Intelligence AI, while using plan generation search techniques derived from both AI action planning and knowledge-based game playing theory. CP1.0, the first version of this planner, is described in this report, and approaches to incorporating advanced planning techniques e.g., metaplanning, hierarchical, and distributed planning are discussed.