This paper has interest in the way to gather many random moving objects and to move of them to target position area. It is not difficult to find random moving objects around us. Crowds in exhibition, cars in urban center, fish in fish farms and other objects which can form groups and change their speed and direction by them can be the example of random moving objects. Objects, which don't have the herd instinct, are not easy to be controlled in a cluster. If the objects move randomly and consist of large numbers, individual control would be one of solutions to move all of them to the desired position. But the individual control might be not only ineffective but also sometimes impossible. Instead of individual object control, it can be solved by gathering whole objects, called clustering, and driving the cluster to move to the target position. To find one of solution, this paper proposes the boundary push method which clusters plural Brownian moving factors and migrates them to the target area such as shepherding sheep in pasture. This paper gives brief introduction of the study at the beginning. In the second chapter, the cluster dynamic modeling to simulate the proposed method is conducted. In the third chapter, boundary push method is proposed and explained. In the forth chapter, the computer simulation is performed to evaluate the proposed algorithm before physical experiment using micro robots.