Planning in Concurrent Domains
Subbarao Kambhampati, Jay M. Tenenbaum · 2004
In many real-world situations, a planner is part of an integrated problem-solving environment and must operate concurrently with other planners and special purpose inference engines. Unlike the traditional AI planners, planners operating in such concurrent environments have to contend with an evolving problem specification, and should be able to interact and coordinate with the other modules on a continual basis. This in turn poses several critical requirements on the planning methodology. In this paper we identify the ability to incrementally accommodate the changes necessitated by the externally imposed constraints into the existing plans, and the ability to understand and reason about the rationale behind externally imposed constraints at an appropriate level of detail as two crucial requirements for planning in such environments. We then explore directions for extending classical hierarchical planning framework to handle those requirements. 1 Introduction In many real-world situat...