. One of the strengths of Partial-Order Causal Link (POCL) planning is the ability to postpone decisions. But because postponed decisions play no role in reasoning about the plan until they are eventually acted upon, the penalty for postponing some decisions can be quite high. We call this style of decision postponement passive postponement, and present experimental results that quantify the efficiency penalty it incurs. We also suggest an alternate approach, active postponement, that allows postponed decisions to impose constraints on the generation of a plan. This constraint-based approach to decision postponement has been implemented in the Descartes planning system. In Descartes, every planning decision is represented by a variable, with constraints on each variable representing criteria that must be satisfied by the corresponding decision. These constraints are managed by a general-purpose constraint engine, so that even postponed decisions play a role in reasoning about the plan....

In every hospital a duty roster for the nursing staff of each ward must be created each month. The roster has to take into account several requirements such as minimal allocation of a station, legal regulations and wishes of the personnel. This planning is very complex and is still done manually. INTERDIP is an advanced industrial prototype that supports semi-automatic creation of such rosters. Using constraint based programming, INTERDIP imitates certain aspects of manual planning. With INTERDIP a roster can be created interactively within some minutes instead of by hand some hours. Additionally, it mostly produces better results. INTERDIP was developed in collaboration with Siemens Nixdorf. It was presented at the Systems'98 Computer exhibition in Munich and several companies have inquired to market our system. 1 Introduction Many real-life problems lead to combinatorial search, computationally a very intensive task. Unfortunately, no general method exists for solving this kind of p...

Expressiveness of constraint programming permits solving elegantly and efficiently many problems but makes difficult for the programmer to debug and understand the behaviour of his programs since most of the solving process (addition of new constraints and propagation of variables’ domains modifications) is concealed by constraint solvers. However, tackling this problem by displaying a graphical representation of the original set of added constraints (store) is useless due to its huge size and its lack of structure whatsoever. We present in this chapter a means to organize hierarchically the store in order to divide it into manageable parts while preserving computation correctness; soundness of the method is shown, an algorithm supporting it is given, and an implemented prototype exhibiting its effectiveness is described. 1

Abstract not found

copies of this thesis and to lend or sell such copies for private, scholarly or scienti c research purposes only. The author reserves all other publication and other rights in association with the copyright in the thesis, and except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatever without the author's prior written permission. Date: