Abstract: The particle swarm optimization algorithm is a new branch in evolution computing field.This algorithm is simple and effective，and is easy in premature convergence.According to the idea of divide and conquer and the enlightenment of co-evolution，two sub-swarms co-evolution particle swarm optimization algorithm is presented.When it is odd number，the algorithm searches and finds optimization in all particles，else particles are divided into two sub-swarms；they implement their own tasks without almost any connection with others.By doing experiments on four benchmark functions，the results show that the algorithm avoids trapping into local optimum in a certain extent and improves the precision of convergence.