To resolve the problem of trapping in a local optimum in particle swarm optimization

a two-stage dynamic cooperative optimization algorithm of multiple particle swarms was proposed.The algorithm contains a master particle swarm and multiple particle swarms.Every particle in the slave particle swarms searchs a part of the problem domain.The master particle swarm coordinates every slave particle swarm converge to the optimum and obtain the optimum searched.In the first stage

the new slave particle swarms will be produced at the problem domain covered few particles

so that the particles can cover the problem domain very well.In the second stage

the slave particle swarms whose positions are overlapping will be deleted in the subregion for decreasing search time.Five benchmark functions are used to compare with two-layer particle swarm optimization.The result illustrate this algorithm can obtain better solutions in optimizing high dimensional multi-peak functions.