This edited book contains papers from the sixth workshop on “planning based on decision theory,” organized at the International Center for Mechanical Sciences (CISM) in Udine, Italy in 2002.

The goal of the workshop was to present some modern approaches to planning and scheduling, in particular approaches developing from the integration of machine learning and operational research. The unifying theme of the book is the decision theory paradigm, an evolution of the tradition in artificial intelligence (AI) to exploit knowledge representation for decision making.

The contributions are short reviews of recent research, sometimes including some general tutorials. The poor integration of the papers is a clear indication that the book is in the proceedings style. All papers were invited, and they are collected into three sessions.

Session 1, “Decision theory,” contains four papers, and presents most of the theoretical work and claims of the book. “Qualitative decision rules under uncertainty,” by Dubois and Fargier, is a long tutorial about what the authors propose to be decision theory, a notation that deals with incomplete knowledge, uncertainty, and utility functions. “Sequential decision making in heuristic search,” by Hullermeier, reviews different heuristics for graph search and change detection, and proposes an iterative method to decide which paths to search. “Identification of non-additive measures from sample data,” by Miranda, Grabish, and Gil, introduces basic concepts in nonadditive measures and presents results in solving optimization problems. “Notification planning with developing information states,” by Schaal and Lenz, is focused on the traveler’s problems. Their representation uses influence diagrams to represent variables and decisions.

Session 2, “Planning, control, and learning,” contains three papers on quite general approaches. “The problem of planning with three sources of uncertainty,” by Traverso, is a review of the planning as model checking approach. “Understanding control strategies,” by Bratko and Suc, presents preliminary results from machine learning techniques to analyze input from the human controllers, and to induce the controller. “Local structure learning in graphical models,” by Borgelt and Kruse, investigates the problem of constructing probabilistic networks.

Session 3, “Application of planning and decision making theory,” contains three papers, on three different application domains in finance and production.

The book is an indication of a growing interest in integrating different approaches and algorithms to solve difficult tasks. As usual in such integrations, the result is at the same time more specialized than needed by the majority of readers, and not broad enough to collect, under the same theoretic view, much of the ongoing research in decision theory.

In conclusion, the book is focused enough to help researchers and graduate students to approach decision theory, and some aspects of the planning domain.