We are interested in solving real-world planning problems and, to that end,
argue for the use of domain knowledge in planning. We believe that the fie
ld must develop methods capable of using rich knowledge models to make plan
ning tools useful for complex problems. We discuss the suitability of curre
nt planning paradigms for solving these problems. In particular, we compare
knowledge-rich approaches such as hierarchical task network planning to mi
nimal-knowledge methods such as STRIPS-based planners and disjunctive plann
ers. We argue that the former methods have advantages such as scalability,
expressiveness, continuous plan modification during execution, and the abil
ity to interact with humans. However, these planners also have limitations,
such as requiring complete domain models and failing to model uncertainty,
that often make them inadequate for real-world problems.
In this article, we define the terms knowledge-based and primitive-action p
lanning and argue for the use of knowledge-based planning as a paradigm for
solving real-world problems. We next summarize some of the characteristics
of real-world problems that we are interested in addressing. Several curre
nt real-world planning applications are described, focusing on the ways in
which knowledge is brought to bear on the planning problem. We describe som
e existing knowledge-based approaches and then discuss additional capabilit
ies, beyond those available in existing systems, that are needed. Finally,
we draw an analogy from the current focus of the planning community on disj
unctive planners to the experiences of the machine learning community over
the past decade.