A Call for Knowledge- Based Planning

Much of the effort of the planning community is currently focused on improving the performance of disjunctive planners (DPs). We are interested in solving real-world planning problems and, to that end, argue for the use of domain knowledge in planning. Hierarchical task network (HTN) planners use more knowledge than DPs and have advantages such as scalability, expressiveness, continuous plan modification during execution, and the ability to interact with humans. We argue, however, that the field must develop methods capable of using even richer knowledge models than those used in HTNs (and therefore DPs) in order to make planning tools useful for complex problems. While we applaud the development of faster DP systems and their use for planning subproblems, it may not be best for the field to focus so many resources on techniques that solve only a narrow subset of the problems that are faced in real-world domains.