georgeslabreche / opssat-orbitai

The OrbitAI app for the OPS-SAT spacecraft.

OPS-SAT OrbitAI

Description

The use of Artificial Intelligence (AI) is of rising interest for space-segment applications despite limited technology demonstrators on-board flying spacecrafts. Past missions have restricted their experience with AI to inferring models that were trained on the ground prior to being uplinked to a spacecraft. The OrbitAI experiment pushes the envelope by breaking away from this trend and shifting ML training from a ground activity to an autonomous in-flight operation. Running on-board the OPS-SAT spacecraft, the experiment uses online Machine Learning (ML) algorithms to train fault detection, isolation, and recovery (FDIR) models that can be used to protect the on-board camera’s lens against exposure to sunlight. One model achieved 89% balanced accuracy in its predictions with an F1 score of 95%.

Citation

We appreciate citations if you reference this work in your upcoming publication. Thank you!

Plain Text

Labrèche, G., Evans, D., Marszk, D., Mladenov, T., Shiradhonkar, V., Soto, T., & Zelenevskiy, V. (2022). OPS-SAT Spacecraft Autonomy with TensorFlow Lite, Unsupervised Learning, and Online Machine Learning. 2022 IEEE Aerospace Conference (AERO), 2022, pp. 1-17, doi: 10.1109/AERO53065.2022.9843402.

BibTex

@INPROCEEDINGS{9843402,
  author={Labrèche, Georges and Evans, David and Marszk, Dominik and Mladenov, Tom and Shiradhonkar, Vasundhara and Soto, Tanguy and Zelenevskiy, Vladimir},
  booktitle={2022 IEEE Aerospace Conference (AERO)}, 
  title={{OPS-SAT Spacecraft Autonomy with TensorFlow Lite, Unsupervised Learning, and Online Machine Learning}}, 
  year={2022},
  volume={},
  number={},
  pages={1-17},
  doi={10.1109/AERO53065.2022.9843402}}

Repository Structure

• Mochi: C++ implementation of online machine learning algorithms.
• RandomForest: C++ implementation of the Random Forest algorithm.
• ipk: Script and guideline to generate a ZIP file and then an IPK file to install the experiment on the Satellite Experimental Processing Platform (SEPP) on-board OPS-SAT.
• nmf/space-app: The OrbitAI NMF App.
• pub: Publication assest, i.e. R scripts, plots, and diagram files.
• results: Mission flight logs, logged training data, and serialized models from the experiment's initial runs on-board the OPS-SAT spacecraft.
• sandbox: A collection of scripts and experiments conducted while the project was being scoped.

Getting Started

1. Initialize and update the Git submodules: git submodule init && git submodule update.
2. Read the READMEs in the nmf/space-app, Mochi, and ipk directories.