A new particle swarm optimization based on the food searching activities of multi-swarm of honeybees

On the basis of analyzing the classical particle swarm optimization (PSO), this paper proposes a new version of the PSO, namely, Honeybee PSO. The Honeybee PSO divides the whole swarm into several small subswarms in which each particle decides its own search direction in the use of roulette. And by this the diversity of the swarm is satisfied. In the process of searching, each particle considers its previously visited best position, the local best position of selective subswarm and its previously visited worst position, which incarnates the `seeking best and avoiding worst´ of the particle and could improve searching efficiency. The algorithm implements the chaotic local search (CLS) according to dimension into the whole best position, which can not only avoid getting into local minimum but also can separate different dimension of the position. By comparing the Honeybee PSO and PSO with two standard testing function, that is GP-Goldstein-Price and RA-Rastrigin, the results show that the Honeybee PSO can hunt out better position and more efficiency than PSO, and so on.