In this paper, we introduce a new disjunctive decomposition scheme for discrete constraint satisfaction problems (CSPs). This strategy is based on first identifying complete no-goods in a graph derived from the microstructure of the CSP, then using these no-goods to decompose the initial CSP into subproblems that exclude these no-goods. This decomposition produces a partition of the solution space and is guaranteed to keep all solutions while reducing the total number of possibilities to be considered. We describe the strategy, study its properties, and identify the number of possibilities that are excluded at any decomposition step. We describe a practical application to which this strategy can be efficiently applied and compare our technique to some decomposition methods reported in the literature.
Constraint Satifaction, decomposition, CSP