The technology of robot formation is constantly developing and gradually being applied. In order to improve the work efficiency of inspection robots, this paper proposes a formation control method based on the leader-follower method and the artificial potential field method, retaining the advantages of the two methods. This allows the robot formation to avoid obstacles while maintaining the formation as much as possible, and improves the local minimum problem that is prone to occur in the artificial potential field method by applying random disturbances. The experimental results show that the time consumed by the proposed method to complete tasks has decreased by 28.5% and 41.4%, respectively, compared to the leader-follower method and the behavior based method. The formation using this method can also perform well in executing specific tasks such as collaborative encirclement and sequential platoon. The feasibility of the method was further verified through simulation of the simulation environment using the ROS (Robot Operating System) platform.