In recent years, due to the shortage of manpower in the security industry, security measures by automatic tracking of camera-equipped drones are expected. The closer a drone is to a moving object, such as a suspicious object, the easier it becomes to identify its features, contributing to surveillance and other operations.

When a drone is brought close to a moving object, it can acquire detailed information, but the possibility of the object being out of the camera's field of view increases because its movements are unknown. On the other hand, if the drone is moved away from the moving object, the moving object can be captured within the angle of view, but the accuracy of the information obtained is lower. Therefore, an optimum positioning control of the drone is required to obtain detailed information while keeping the moving object within the angle of view of the camera.

To achieve this, it is necessary to estimate the future movements of the moving object and control the drone based on the estimated values. However, there is always an error between the actual and the estimated movement of the mobile object, which may cause the mobile object to deviate from the camera's angle of view. Existing methods have not considered drone tracking control that takes the estimation error into account.

The aim of this research is to control drone tracking according to the accuracy of the estimation of the moving object's movement. In order to achieve this objective, we propose an optimal drone positioning control using Gaussian process regression to estimate the position of the moving object and its accuracy, and model predictive control, and clarify its effectiveness through simulations and experiments.