Online Supervised Training of Spaceborne Vision during Proximity Operations using Adaptive Kalman Filtering

TL;DR

This paper introduces an online supervised training method that uses adaptive Kalman filtering to improve vision-based navigation for spacecraft during proximity operations, addressing the domain gap issue in neural networks trained on synthetic data.

Contribution

It proposes a novel online training approach that updates a pose estimation neural network during flight using adaptive Kalman filter pseudo-labels, enhancing real-world performance.

Findings

OST improves neural network accuracy on real target images.

Training with diverse target views enhances performance.

Method effectively bridges the synthetic-to-real domain gap.

Abstract

This work presents an Online Supervised Training (OST) method to enable robust vision-based navigation about a non-cooperative spacecraft. Spaceborne Neural Networks (NN) are susceptible to domain gap as they are primarily trained with synthetic images due to the inaccessibility of space. OST aims to close this gap by training a pose estimation NN online using incoming flight images during Rendezvous and Proximity Operations (RPO). The pseudo-labels are provided by adaptive unscented Kalman filter where the NN is used in the loop as a measurement module. Specifically, the filter tracks the target's relative orbital and attitude motion, and its accuracy is ensured by robust on-ground training of the NN using only synthetic data. The experiments on real hardware-in-the-loop trajectory images show that OST can improve the NN performance on the target image domain given that OST is performed on images of the target viewed from a diverse set of directions during RPO.