Automatic relative pose control of unmanned aerial vehicles based on diversified point cloud matching algorithm

Drones are prone to getting stuck in a single extremum point in the neighboring coordinate system of the flight environment in complex and changing environments, which increases the horizontal and vertical disturbances in the pose matching and control process, resulting in poor position and attitude tracking accuracy. To address this, an automatic control method based on diversified point cloud matching algorithm is proposed. By integrating depth information and curvature for point cloud regionalization, introducing neighborhood and curvature to calculate covariance matrices, a new feature calculation formula is obtained. With the representation ability of the new feature, the tracking differentiator and extended state observer are automatically optimized and designed, and a nonlinear error feedback control law is obtained to automatically control the relative pose of the drone. The experimental results show that the proposed method for point cloud matching has a horizontal and vertical error of only 0.04m-0.15m. In terms of relative pose control, the drone pose can be effectively controlled under various conditions such as no disturbance, instantaneous disturbance, and white noise disturbance.