Polaris: Machine Learning for cubesat awareness and diagnostic
Polaris: Machine Learning on health-keeping telemetry for cubesat awareness and diagnostic
version 0.3
Aim
The aim is to analyze a satellite set of telemetry to understand links/dependencies among different subsystems. The project should be able to demonstrate understanding of the links between the different behaviour changes of each telemetry within a satellite or within a set of external sources of information (mission plan, solar aspect angles, ephemerides, etc.). Machine learning can be used to learn the different link models and storage of acquired knowledge should be stored in a graph (Bayesian network). The intermediate and final output should be represented as data interpretable by a visualization interfaces, preferably in JSON.
Skills/Knowledge required
- Python programming
- Pandas data wrangling
- Feature engineering principles
- Scikit-learn pipelines
- Traditional machine learning workflow as well as continuous learning workflow
- ML algorithms and models: xgboost, auto-encoders, generative models
Expected results
The final contribution should be a python module with an usable API on pandas dataframes (or eventual datastreams)
Potential mentor(s)
- Redouane Boumghar: Data scientist for space ops
- Xabier Crespo Alvarez: Embedded System Consultant
- Hugh @SaintAardvrak: Project management, SatNOGS data
Related repositories
Explorations and tests already available here: https://gitlab.com/crespum/polaris/
Wiki pages