00525294

b

00525294 Knoblock, Craig A. Generating abstraction hierarchies. An automated approach to reducing search in planning. The Kluwer International Series in Engineering and Computer Science. 214. Dordrecht: Kluwer Academic Publishers. xi, 168 p. Dfl. 157.50; US\$ 80.00; \sterling 58.00 /hc (1993). 1993 Dordrecht: Kluwer Academic Publishers EN abstraction hierarchies operator-based problem solving This book presents a fully automated approach to generating abstraction hierarchies in operator-based problem solving. The abstractions are generated by a domain-independent algorithm whose inputs are the definition of a problem space and the problem to be solved and whose output is an abstraction hierarchy that is tailored to the particular problem. This algorithm is implemented in the system ALPINE. It generates abstractions for a hierarchical version of the PRODIGY problem solver that was developed at Carnegie Mellon University. The generated abstractions are tested in several domains on large problem sets. They produce shorter solutions with significantly less search step than problem solving without using abstractions. The author creates the notion of the ``ordered monotonicity'' property of an abstraction hierarchy which guarantees that the structure of an abstract solution is not changed in the process of refining it. Abstraction hierarchies that enjoy this property allow for considerable reductions in planning time. The book is written in a nice and readable style. It contains several illustrations and examples. It is an updated version of the author's doctoral thesis and it starts with a foreword written by Jaime Carbonell. J.Desel (Berlin)