Research On Bee-Swarm Genetic Algorithm

Genetic Algorithm was proposed in 1970s by professor Holland. It is an effective optimization method to stimulate the evolution of nature. Based on the accumulated knowledge, GA can not only globally search the unknown space by probability searching mechanism, but also change by the different problem, so it is widely applied. However, the traditional GA has the disadvantage of dealing with the slower rate and premature convergence, causing by the contradiction between population diversity and pressure of selection. So many different methods are proposed to balance their contradiction. Although a lot of improved GAs make progress, it still has some limitations.Bee-Swarm genetic algorithm based on reproducing of swarm is proposed in this thesis, to solve the contradiction between population diversity and pressure of selection. The swarm is composed of queen bee, drones and worker bees, the main operators of the algorithm include the absolute mating right between queen bee and drones, the simulated suppression between queen bee and worker bees, local optimization of queen bee based on simulated annealing, adaptive crossover between drones and queen bee, adaptive mutation of worker bee. Then the convergence and validity are proved by the analysis of Markov mode and genetic mechanism. Finally, the simulation results about the optimization of functions and combinational optimization about N-queen problem show that it has both validity and less computational magnitude.