Deep Learning-Based Target Pose Estimation Using LiDAR Measurements in Active Debris Removal Operations

In this article, a study on the use of a commercial global-flash LiDAR sensor in active debris removal operations is presented. This type of activity requires precise knowledge of the position and orientation of the target to be removed. For these missions, relative navigation devices, such as cameras or LiDAR sensors are typically regarded. In this study, the mission profile defined in the e.Deorbit System Requirements Review was considered, and data acquisition and processing from a commercial ASC GSFL-16KS LiDAR sensor were simulated. As the main novelty of this work, the use of multilayer perceptron neural networks for the processing of LiDAR depth images is proposed in order to obtain an estimate of the pose of the target. Using the results of the neural networks, an iterative closest point algorithm is applied to refine the calculation of the pose. The accuracy and computation time of the system were evaluated, obtaining robust and computationally efficient results in the proposed study cases.

active debris removal (adr); global-flash lidar; multilayer perceptron (mlp); pose estimation

Deep Learning-Based Target Pose Estimation Using LiDAR Measurements in Active Debris Removal Operations Articles