LiDAR-Based Pose Initialization Dataset for Spinning Spacecrafts
Abstract
This dataset contains fully synthetic LiDAR data generated from scratch and used in the article: Recurrent Convolutional Neural Networks for LiDAR-Based Pose Initialization of Spinning Spacecrafts Authors:Luca Bechis, Jean-Luc Sarvadon, Petre Ricioppo, Mauro ManciniDepartment of Mechanical and Aerospace Engineering, Politecnico di Torino, Italy Dataset Description The dataset is designed for LiDAR-based pose initialization and relative navigation of spinning spacecraft. It includes two main components:1) Image Dataset: image-based LiDAR representations used as inputs for convolutional neural networks.2) Sequence Dataset: temporal sequences of LiDAR scans used for recurrent neural network training. The dataset includes data generated for three spacecraft models:- Aquarius- Magellan- NEAR Shoemaker The data are fully synthetic and generated using a custom simulation pipeline. Citation If you use this dataset in your work, please cite both this dataset and the associated article. @article{Bechis2025RCNNLiDAR, title = {Recurrent Convolutional Neural Networks for LiDAR-Based Pose Initialization of Spinning Spacecrafts}, author = {Bechis, Luca and Sarvadon, Jean-Luc and Ricioppo, Petre and Mancini, Mauro}, journal = {IFAC World Congress}, year = {2025}, note = {Under review} Dataset Repository The dataset repository, including future releases of the dataset generation code, is available at:https://github.com/STREAMRobotics/rcnn-lidar-spinning-spacecraft-dataset