A genetic algorithm for the capacitated minimum spanning tree problem

This work describes a Genetic Algorithm for the Capacitated Minimum Spanning Tree (CMST) problem that appears in Telecommunications. In the CMST problem, the tree is composed by a set of rooted subtrees with limited number of nodes due to capacity constraints. We suggest a new crossover operator, applied directly to the tree (phenotype) instead of to the chromosome. The crossover works by transfering nodes among rooted subtrees. The new crossover proved to be effective when compared with the Genetic Algorithm of Raidl and Drexel (on benchmark data sets) in which presented better performance than various classical methods of the literature.