Abstract: To apply the structural persistence graph in the formation of dual-wheel robots, unmanned aerial vehicles (UAVs), and multi-robot systems, as well as the visual simultaneous location and mapping (SLAM) algorithm for positioning, we study the formation method, which combines minimally persistent formation with potential field control. The SLAM algorithm provides robots with accurate position information. First, minimally persistent formation is generated based on the improved directed increasing vertex operations. Secondly, we propose a visual SLAM algorithm based on an RGB-D camera to obtain the robot's exact position. Thirdly, we design a potential field function for multi-robots in the formation. Then, we propose a two-wheeled robot formation control method based on the potential field function that considers the non-integrality of two-wheeled robots. The position information used in the control process is determined by the visual SLAM algorithm. Then, we realize the minimally persistent formation of multi-robots in different planes combined with UAVs and multi-robot systems. Lastly, we verify the validity of the proposed method on the Gazebo simulation platform.