Machine Learning in Deep Space

Abstract: Applying machine learning on-board deep space missions presents a unique set of technical and cultural challenges to the space industry. Terrain assessment is a essential process for autonomous rovers to be able to safely navigate their way across unstructured terrain which we use as a case study for machine learning in this sector. A new software tool was developed to be able to deploy this and other ML models onto the radiation hardened processors required on deep space missions. This tool was required technically since the Sparc V8 architecture used wasn't supported and because of the need to adhere to the strict coding standards used in space. Finally the verification and validation challenges and the changes needed within systems engineering culture before these solutions will actually fly will be discussed.

Bio: Pete Blacker is a PhD student working at the Surrey Space Centre in Guildford, sponsored by Airbus he specialises in perception and autonomy on-board Mars rovers. Originally studying computer science at the University of Manchester he went on to complete a masters in space engineering at the University of Surrey before starting his PhD.