LiDAR based pose tracking of an uncooperative spacecraft using the smoothed normal distribution transform

Kurzfassung

Lidar sensors provide precise 3D point cloud measurements, and can be used for pose estimation of an uncooperative satellite during space rendezvous. For updating the estimated pose of a the target, iterative closest point (ICP) or one of its variants is usually applied as a tracking method. However, for dense point clouds and space hardware with reduced computing power, the execution time of ICP can become a limiting factor. Normal distribution transform (NDT) is an alternative algorithm for fine point cloud registration, which can be faster than ICP. Yet, NDT can be less robust than ICP due to the discontinuities in its cost function. This work proposes a smoothing method of the NDT map in order to mitigate this robustness problem. In addition, a strategy for correcting motion blur observed with lidar sensors when recording a tumbling target is developed. Experiments at the European Proximity Operations Simulator demonstrate the efficiency, precision and robustness of the smoothed NDT algorithm when compared to ICP, as well as the importance of motion blur correction for precise pose estimation.