Several attempts have been made to enhance PSO performance by combining it with a local search method. Following the same track, we present in this paper local search in PSO performed by smaller independent swarms of PSO producing PSO2. Different modifications are made to help basic PSO2 enhance performance. PSO2-RS and PSO2-SA are 2 modified versions of PSO2 that targeted to increase the swarm diversity. Increasing the local search swarms sizes as the search progresses is another modification made to basic PSO2 in order to change the algorithm behavior to be more exploitive. The final algorithm is examined against 4 functions of the CEC-2005 benchmark suite and results are reported.