In AI planning research, planning practice (as embod-ied in implemented planning systems) tends to runfar ahead of the theories that explain the behavior ofthose planning systems. For example, the past fewyears have seen much analysis of the properties oftotally- and partially-ordered planning systems usingSTRIPS-style planning operators (Minton et al., 1991;McAllester and Rosenblitt, 1991; Chapman, 1987).However, most of the practical work on AI planningsystems for the last fifteen years has been based onhierarchical task-network (HTN) decomposition(Sac-erdoti, 1990; Tate, 1977; Wilkins, 1984) as opposed toSTRIPS systems, yet there has been very little simi-lar analytical work on the properties of hierarachicaltask-network (HTN) planning systems.One of the primary obstacles impeding such workhas been the lack of a clear theoretical framework ex-plaining what a HTN planning system is. A primarygoal of our current work is to correctly define, analyze,and explicate features of the design of HTN planningsystems. In this report, we describe some first stepstoward that goal: We set out a formal definition ofHTN planning, present a nondeterministic HTN plan-ning procedure P
[1]
Earl D. Sacerdoti,et al.
The Nonlinear Nature of Plans
,
1975,
IJCAI.
[2]
Dana S. Nau,et al.
Semantics for hierarchical task-network planning
,
1994
.
[3]
David Chapman,et al.
Planning for Conjunctive Goals
,
1987,
Artif. Intell..
[4]
David E. Wilkins,et al.
Domain-Independent Planning: Representation and Plan Generation
,
1984,
Artif. Intell..
[5]
David A. McAllester,et al.
Systematic Nonlinear Planning
,
1991,
AAAI.
[6]
Austin Tate,et al.
Generating Project Networks
,
1977,
IJCAI.