A comparison of multiobjective evolutionary algorithms with informed initialization and kuhn-munkres algorithm for the sailor assignment problem

This paper examines the performance of two multiobjective evolutionary algorithms, NSGA-II and SPEA2, with informed initialization on large instances of United States Navy's Sailor Assignment Problem. The informed initialization includes in the initial population special solutions obtained by an extension of the Kuhn-Munkres algorithm. The Kuhn-Munkres algorithm, a classical algorithm that solves in $O(n^3)$ time instances of the single valued linear assignment problem, is extended here to render it applicable on single objective instances of the sailor assignment problem obtained using weight vectors to scalarize the natural multiobjective formulation. The Kuhn-Munkres extension is also used to provide a performance benchmark for comparison with the evolutionary algorithms.