Mobile-URSONet: an Embeddable Neural Network for Onboard Spacecraft Pose Estimation

TL;DR

This paper introduces Mobile-URSONet, a lightweight neural network designed for onboard spacecraft pose estimation, achieving significant parameter reduction with minimal accuracy loss to enhance autonomous space operations.

Contribution

The paper presents Mobile-URSONet, a novel neural network architecture that drastically reduces parameters while maintaining high accuracy for spacecraft pose estimation.

Findings

178 times fewer parameters than URSONet

Degrades accuracy by no more than four times

Suitable for onboard spacecraft deployment

Abstract

Spacecraft pose estimation is an essential computer vision application that can improve the autonomy of in-orbit operations. An ESA/Stanford competition brought out solutions that seem hardly compatible with the constraints imposed on spacecraft onboard computers. URSONet is among the best in the competition for its generalization capabilities but at the cost of a tremendous number of parameters and high computational complexity. In this paper, we propose Mobile-URSONet: a spacecraft pose estimation convolutional neural network with 178 times fewer parameters while degrading accuracy by no more than four times compared to URSONet.