This paper attempts to discuss the following problems that must be solved to realize the return to the original position: 1, attitude control of the aircraft itself, 2, flight trajectory memory, 3, autonomous navigation. First of all, the attitude control technology of the aircraft itself is very mature, and the second point, flight trajectory memory, is very complicated to realize and must be simplified into an equivalent model. The starting point (origin) is A and the farthest point of throwing. Then the whole trajectory storage space is a three-dimensional array of circles with radius ab and height H, that is to say, there are three parameter sets of radial and axial * * * *, so at least three sensors are needed to provide data streams for the processor to "judge" and finally calculate the coordinates of point A and point B. In fact, the processor has no wisdom to judge, but only executes the program of flight control software. Firstly, the steady-state change is used as an event to trigger the orientation sensor to work and restore the interruption of the steady-state sensor. The positioning program calculates the flight trajectory through three sets of azimuth sensor data, one set of speed sensor data and clock frequency, and then returns to the program to control the plane to fly back to the origin. The whole process should be 1, 2, 3, and executed concurrently. Of course, the specific process is complicated and troublesome, and even some bottlenecks cannot be solved for a long time, but it doesn't matter. For flight enthusiasts, they enjoy the ideal process.