Dynamic Segregative Genetic Algorithm for Assembly Lines Balancing

This paper presents a segregative genetic algorithm for "I"/U"-shaped assembly line balancing problem. It uses a basic genetic algorithm and a feature function that associates a time profile of the workstations to each chromosome. The similarity based clustering in the feature space induces subpopulations of chromosomes. The segregative genetic algorithm acts both on representation and feature space. A similarity based communication preserves the clustering structure. Each subpopulation completely exploited sends its centroid to an associative tabu search mechanism. Some selected new individuals are used to create clusters that represent unexplored parts of search space. The exhausted subpopulations are replaced by new ones during the evolution. The resulted dynamic segregative genetic algorithm leads to a better trade-off between exploration, made by many clusters, and exploitation, done by the focusing on each subpopulation. Experimental investigations show that the segregative approach is more stable and systematically produces better results than the basic genetic algorithm. A distributed implementation of the segregative approach is presented and its performance is reported.