Aiming at the formation problem of unmanned aerial vehicles (UAVs), a relative positioning method between UAVs is proposed. Based on the high precision carrier phase measurement technology and the extended Kalman filter (EKF) algorithm, the synchronous solution of the position, velocity and attitude of the UAV is realized by using three radio transceivers configured in a triangle on each UAV; The relative positioning model and the specific filtering algorithm are given, and the flight States of the UAV group in different spatial layouts such as line, “surface” and “body” are estimated and analyzed. The results show that compared with the “line” layout, the “surface” layout can significantly improve the positioning performance, and there is no significant difference with the “body” layout, which indicates that in the UAV group positioning system, the rational configuration of observation geometry can effectively improve the relative positioning accuracy. Simulation examples show that under the condition of good observation geometry, redundant observation data can further improve the system stability.