Adaptive Neural Network-based Unscented Kalman Filter for Robust Pose Tracking of Noncooperative Spacecraft

This paper presents a neural network-based Unscented Kalman Filter (UKF) to estimate and track the pose (i.e., position and orientation) of a known, noncooperative, tumbling target spacecraft in a close-proximity rendezvous scenario. The UKF estimates the relative orbit and attitude of the target with respect to the servicer based on the pose information provided by a multi-task Convolutional Neural Network (CNN) from incoming monocular images of the target. In order to enable reliable tracking, the process noise covariance matrix of the UKF is tuned online using adaptive state noise compensation. This is done through a newly developed process noise model for relative attitude dynamics in closed form. In order to enable a comprehensive analysis of the performance and robustness of the proposed CNN-powered UKF, this paper also introduces the Satellite Hardware-In-the-loop Rendezvous Trajectories (SHIRT) dataset. SHIRT comprises the labeled images of two representative rendezvous trajectories in low Earth orbit created from a graphics renderer and a robotic testbed. Specifically, while the CNN is solely trained on data from computer graphics, the functionality and performance of the complete navigation pipeline are evaluated on actual Hardware-In-the-Loop (HIL) images from the robotic testbed as well. The proposed UKF is evaluated on SHIRT's synthetic and HIL images and is shown to have sub-decimeter-level position and degree-level orientation errors at steady-state for separations less than 10 meters.