Autonomous navigation around didymos using CNN-based image processing

Kaluthantrige, Mewantha and Fodde, Iosto and Feng, Jinglang and Gil-Fernández, Jesús (2022) Autonomous navigation around didymos using CNN-based image processing. In: AAS/AIAA Astrodynamics Specialist Conference 2022, 2022-08-08 - 2022-08-11.

[thumbnail of Kaluthantrige-etal-AAS-AIAA-2022-Autonomous-navigation-around-didymos-using-CNN-based-image-processing]
Text. Filename: Kaluthantrige_etal_AAS_AIAA_2022_Autonomous_navigation_around_didymos_using_CNN_based_image_processing.pdf
Accepted Author Manuscript
License: Strathprints license 1.0

Download (2MB)| Preview


HERA is an asteroid rendezvous mission of the European Space Agency (ESA) that will investigate the binary asteroid system (65803) Didymos. The proximity operations of HERA's spacecraft around the target body rely on an autonomous optical navigation system that collects on-board visual information to estimate the relative position and attitude of the spacecraft with respect to the asteroid. The core component of this navigation method is the Image Processing (IP) algorithm that extracts optical observables from images captured by the spacecraft's on-board Asteroid Framing Camera (AFC). The Early Characterization Phase (ECP) is a proximity operation of HERA with the objective of conducting physical and dynamical characterizations of Didymos. This paper develops a pipeline to estimate the position of HERA spacecraft around binary asteroid system Didymos during the ECP using a Convolutional Neural Networks (CNN)-based IP algorithm. The proposed algorithm uses the images captured with the AFC camera to estimate the pixel position of the centroid of Didymos and the pixel position of a set of keypoints on the visible border of Didymos. With these points, the algorithm evaluates the apparent radius of the primary, which is used to measure the pseudorange by applying the pinhole camera model. Subsequently, the algorithm combines the measured centroid and pseudorange with an Unscented Kalman Filter (UKF) to estimate the relative position of the spacecraft. The training, validation and testing datasets are generated with the software Planet and Asteroid Natural scene Generation Utility (PANGU). The High-Resolution Network (HRNet) is used as CNN architecture as it represents the state-of-the-art technology in keypoint detection. The HRNet-based IP algorithm measures the pseudorange and estimates the position of the centroid with high accuracy. The UKF is able to process the centroid and pseudorange measurements to produce an accurate state estimate with an error of around 300 m and 2 cm/s.