Modelling CSP Solution Algorithms with Petri Decision Nets

The constraint paradigm provides powerful concepts to represent and solve different kinds of planning problems, e. g. factory scheduling. Factory scheduling is a demanding optimization task. Typically a large and conflicting set of restrictions, objectives and preferences has to be considered. Real scheduling problems are too complex for simple solution algorithms like backtracking. Instead, problem specific heuristics or expert knowledge must be integrated into the algorithms. In the first part of the paper, a unified formalism is introduced for representation and implementation of complex CSP solution algorithms. The formalism enables the consideration of the problem structure as well as given objectives and the usage of problem specific heuristics or expert knowledge. It is based on the fundamental assumption that designing CSP solution algorithms means designing decision networks. For the class of planning problems considered, we use Petri decision nets (PDN) for representing the ...

• Type: Conference
• Year: 1998
• Where: CP
• Authors: Stephan Pontow
• Keywords: Artificial Intelligence, CP 1998, CSP Solution Algorithms, Factory Scheduling, Solution Algorithms