Genetic algorithms for multicast routing with multiple QoS requirements

In this paper we explore the use of genetic algorithms (GAs) for solving the multicast routing (MCR) problem with multiple quality of services (QoS) constrains. We study the impacts of encoding, initialization, repair, crossover, and selection on the performance of the GAs. An experimental design, with 32 cells and 18 data sets with 10 repetitions in each cell, is used for the empirical tests. The results indicate that encoding, selection, and repair/penalty have significant impacts on solution quality. Among the various options, the combination of determinant encoding, heuristic initialization, uniform crossover, repair, and fittest selection often provides better solution qualities than other combinations. Using this combination, the proposed GA not only obtains solutions significantly better than a shortest path based heuristic, its quality is also robust and almost the same as optimal solutions.