This paper presents EMS , an implemented HTN planning algorithm using a novel form of abstraction. A plan is viewed as a set of dynamic objects taking part in sequences of transitions. EMS builds up plans using an Expand then Make-Sound cycle: a plan is built up by constructing and expanding a tree of networks at different levels of abstraction, where a network contains object transition sequences. A network is made sound by adjusting the pre and postconditions of all the object transition sequences it contains, and proving the sequence in between is sound. As new objects are discovered in the detailed levels of the hierarchically developing plan, the pre and postconditions of their transition sequences are passed up to a parent network in the hierarchy which must then be made sound. The main benefits of EMS are that it processes an expressive, declarative input language based on object hierarchies; it is efficient in its reasoning, in that object transition sequences can be manipulated independently of objects of unrelated sorts, and reasoning about condition achievement is performed locally in a network; and it provides a clear, sound algorithm with the potential of application to complex applications.

Completion of partial plans is a subtask for many planning techniques such as plan reusing, replanning and accomplishing complex user goals. The new generation of fast planners such as Graphplan, Satplan and others, is characterized by very efficent planning algorithms which exploit techniques of multiple plans representation. Unfortunately it is fairly difficult to give Graphplan a partial plan fragment to complete, because the internal plan network representation, only represents ordered by levels and complete plans. On the other hand it is very easy and straigthforward to provide an initial partial plan to complete to a planner like UCPOP, because its planning algorithm completes a partial initial plan in the internal representation. We present an original technique for using Graphplan in order to solve partial plan completion problems. The technique, called domain embedding, modifies the problem domain in order to induce only solutions which are completion of the given partial pl...

In this paper we argue generally for a move to an object-centred formulation of precondition planning. Our previous work has demonstrated some of the advantages of systematically creating and compiling an object-centred domain model. Here we describe a complementary object-centred planning engine which exploits such a domain model, and review some of its benefits. We also survey related work that has attempted to exploit the object-oriented approach in planning. 1 Introduction Much AI planning research has concentrated on the design, analysis and efficiency of planning algorithms, with relatively little attention being paid to the creation and capture of planning domain models. Representational issues can have a big impact on planner performance. For example, it has been shown that different encodings of the same problem domain used with the same planner can give different results [9]. The construction of planning domain models is an important issue in AI planning, especially if we wi...

In considering real applications, one of the major problems in the application of generative planning is the poor performance of general planning...