15th European Conference on Artificial Intelligence
|July 21-26 2002 Lyon France|
In recent years, it is increasingly recognised that action planning in real-world domains requires an accurate treatment of uncertainty. The theory of partially-observable Markov decision processes has been found to provide a powerful framework for studying this type of planning. Within this framework, plans are often expressed as rooted trees. However, for various reasons it is often more convenient to express plans as collections of decision rules. For instance, domain experts are often able to formulate a number of reliable decision rules that could serve as a starting point in finding an optimal plan. This paper investigates the representation of decision-theoretic plans as sets of symbolic decision rules. It is shown under which conditions such plans are internally consistent, coherent, en complete.
Keywords: Planning, Reasoning under Uncertainty, Knowledge Representation
Citation: Niels Peek: Representation of decision-theoretic plans as sets of symbolic decision rules. In F. van Harmelen (ed.): ECAI2002, Proceedings of the 15th European Conference on Artificial Intelligence, IOS Press, Amsterdam, 2002, pp.591-595.