An Elegant and Efficient Implementation of RDS for VWCSP

The Russian Dolls Search (RDS) algorithm is both a solving procedure and a cost based filtering algorithm suitable for Weighted Constraint Satisfaction Problems (WCSP). In the context of Variable Weighted CSP, we propose a declarative implementation of RDS filtering, using a set of auxiliary variables and constraints. This approach allows relying on a Constraint Programming (CP) solver instead of developing dedicated machinery. We prove that this model yields strictly better propagation than the standard one and we bring experimental results showing that both computation time and number of explored nodes are significantly decreased.