摘要：The classical set partitioning problem is defined on a graph G = (V, E) and consists of partitioning V in disjoint subsets. In this study, we propose a Split procedure to minimize the number of partitions for a given sequence of nodes, considering local and global capacity constraints on the partitions. The procedure relies on dynamic programming and it is applied here to the SONET Ring Assignment Problem. It is used as a decoding function in a Biased Random Key Genetic Algorithm and as a constructive heuristic in a hybrid multistart Evolutionary Local Search. High-quality results are computed on classic benchmarks.
