Pose Estimation Techniques for Non-Cooperative Spacecraft Navigation

Accurate determination of an uncooperative spacecraft’s position and orientation—its six‐degree‐of‐freedom (6DOF) pose—is essential for tasks such as active debris removal, on-orbit servicing and autonomous docking. Traditional model-based methods employ LIDAR-derived point clouds and iterative algorithms to align known target geometries, while vision-based strategies use single or stereo cameras to infer pose from image features. More recently, deep learning approaches have enabled direct regression of pose parameters or tight coupling of convolutional feature detectors with classical solvers and filters. To ensure reliability, novel testbeds replicate space lighting and dynamics on the ground, and unsupervised domain-adaptation techniques mitigate gaps between synthetic training data and real imagery. Onboard implementation using hardware-accelerated inference makes neural methods viable in resource-limited environments. Together, these advances form a robust toolset for non-cooperative spacecraft navigation, addressing variable lighting, unknown target models and computational constraints.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Pose Estimation Techniques for Non-Cooperative Spacecraft Navigation publication trend

The graph below shows the total number of articles in pose estimation techniques for non-cooperative spacecraft navigation across all publications each year (not limited to Nature Index journals).

Technical terms

6DOF (Six Degrees of Freedom): The combined translational and rotational motions defining a rigid body’s pose in three-dimensional space.

Monocular pose estimation: The inference of a target’s 6DOF pose from images captured by a single camera.

Perspective-n-Point (PnP) problem: A geometric method to recover camera orientation and position from known 3D points and their 2D image projections.

Domain shift: The degradation of algorithm performance when models trained on synthetic or lab data are applied to real-world imagery with different characteristics.

Iterative Closest Point (ICP): An algorithm that refines relative pose by minimising distances between corresponding points in two overlapping 3D point clouds.

References

1. On-ground validation of orbital GNC: Visual navigation assessment in robotic testbed facility. Astrodynamics (2024).
2. Point Cloud Based Relative Pose Estimation of a Satellite in Close Range. Sensors (2016).
3. A Model-Based 3D Template Matching Technique for Pose Acquisition of an Uncooperative Space Object. Sensors (2015).
4. Vision-Based Spacecraft Pose Estimation via a Deep Convolutional Neural Network for Noncooperative Docking Operations. Aerospace (2020).
5. Evaluation of tightly- and loosely-coupled approaches in CNN-based pose estimation systems for uncooperative spacecraft. Acta Astronautica (2021).
6. On-ground validation of a CNN-based monocular pose estimation system for uncooperative spacecraft: Bridging domain shift in rendezvous scenarios. Acta Astronautica (2022).
7. Utilization of FPGA for Onboard Inference of Landmark Localization in CNN-Based Spacecraft Pose Estimation. Aerospace (2020).
8. Stereo‐Vision‐Based Relative Pose Estimation for the Rendezvous and Docking of Noncooperative Satellites. Mathematical Problems in Engineering (2014).
9. A survey on deep learning-based monocular spacecraft pose estimation: Current state, limitations and prospects. Acta Astronautica (2023).
10. Bridging the Domain Gap in Satellite Pose Estimation: A Self-Training Approach Based on Geometrical Constraints. IEEE Transactions on Aerospace and Electronic Systems (2023).