Robust On-Manifold Optimization for Uncooperative Space Relative Navigation with a Single Camera

Abstract

Optical cameras are gaining popularity as the suitable sensor for relative navigation in space due to their attractive sizing, power and cost properties when compared to conventional flight hardware or costly laser-based systems. However, a camera cannot infer depth information on its own, which is often solved by introducing complementary sensors or a second camera. In this paper, an innovative model-based approach is demonstrated to estimate the six-dimensional pose of a target relative to the chaser spacecraft using solely a monocular setup. The observed facet of the target is tackled as a classification problem, where the three-dimensional shape is learned offline using Gaussian mixture modeling. The estimate is refined by minimizing two different robust loss functions based on local feature correspondences. The resulting pseudo-measurements are processed and fused with an extended Kalman filter. The entire optimization framework is designed to operate directly on the SE(3) manifold, uncoupling the process and measurement models from the global attitude state representation. It is validated on realistic synthetic and laboratory datasets of a rendezvous trajectory with the complex spacecraft Envisat, demonstrating estimation of the relative pose with high accuracy over full tumbling motion. Further evaluation is performed on the open-source SPEED dataset.

ChiNet: Deep Recurrent Convolutional Learning for Multimodal Spacecraft Pose Estimation

“…Generally, global features (e.g. bags of keypoints, shapes, or even raw images) have been preferred for combination with a variety of ML techniques ranging from nearest neighbour search [26] to unsupervised clustering [27], principal component analysis [28], Bayesian classification [29], and deep learning [30].…”

Using Convolutional Neural Networks for Relative Pose Estimation of a Non-Cooperative Spacecraft with Thermal Infrared Imagery

“…We utilise the same concept as Rondao et al [5] of viewsphere sampling which is taken from 3D object detection. This concept imagines a sphere around the target, the surface of which is covered in cameras pointed toward the target at the centre, as can be seen in Fig.…”

“…We have chosen to focus on a simpler problem statement which is to provide an initial estimation of the pose so that a more precise algorithm can make refinements. This had been previously achieved by Rondao et al [5] where a Bayesian classifier, a traditional machine learning algorithm, was used to match 2D features from a visible camera feed to keyframes of the 3D target that were learned offline. This method does automatically extract feature from an image but these features must be determined by the developer beforehand.…”

Optimized Design of 3D Spatial Images Based on Kalman Filter Equation

“…By imposing this constraint on the pathological problem, the problem is subsequently transformed into a fitness problem. Our proposed fast feedback model is also an improved example of a regularized constrained cost function as shown in equation (6). The λ is a weighting factor which is used to balance the two constraint terms to prevent disjoint minimization.…”

“…This article attempts to focus on the interaction between viewers based on the Kalman filter equation and three-dimensional space. This research can also enable image art works to better interact with viewers and artists, so that viewers can have a better and deeper experience of three-dimensional images and can also better participate in three-dimensional images, optimizing design [6,7].…”