Considerable Factors for Domain Model Learning in Planning

Anonymous

Keywords: Knowledge Engineering for Planning, Domain Model Learning, Planning Paradigms

Abstract:

Learning for planning has extensive aspirations to aim at. Our primary focus area of review in this paper is the knowledge engineering of domain models to feed planners. This concerns with learning and representation of knowledge about operator schema, discrete or continuous resources, processes and events involved. With the increased problem complexity on the continuum of planning from classical towards full scope, synthesizing expressive and intensive planning domain models by hand become more challenging, time-intense and error-prone. Our intended contribution is to provide a broader perspective on the range of research in the domain model learning area and specification of interdependent AP dimensions which can affect the developmental decisions of learning methods/techniques.
We map the amplitude of recently created domain model learning methods on five considerable dimensions. The dimensions to underpin the development of a learning system include the type of planning paradigm, representation languages, learning at various planning stages, learning systems and sources and the extent of learning that takes place.
We also identify considerable issues for future work.