DIN Standards Committee Aerospace
Space engineering - Machine Learning Qualification for Space Applications Handbook
Abstract
AI will be a major driver for raising space systems autonomy especially for future exploration activities but also for reducing operations costs of LEO systems and for robotic elements. Today AI cannot be verified and validated for critical applications as the ECSS requirements on software quality are not adapted to be applied for learning systems but for deterministic control algorithms. Likewise, there are a multitude of potential use-cases for AI within the space domain, e.g. health monitoring, AOCS, VBN, image processing and more, but the qualification of such developments has to consider application specific constraints. Therefore the justification of the document is to give guidelines in a handbook on how to qualify ML models for different kind of space software projects while being compliant to the reference standards ECSS-E-ST-40C and ECSS-Q-ST-80C. The scope of the document includes: 1. the recall of the major elements of the machine learning approach (e.g. different kind of topologies and methods) together with best practices for development and verification. 2. the “appropriateness” of each of these approaches for various kind of projects (flight & ground autonomy, link to situational awareness, assistance systems, etc.), including methods of validation associated to application use-case. 3. the mapping of these practices on the relevant sections of ECSS-E-ST-80C and ECSS-E-ST40. 4. the proposal of guidelines to implement Standards' requirements for the given practice (e.g. how to handle testing and overall qualification). The scope of the document is not: • to propose a risk based tailoring of the ECSS standards. • to propose contractual provisions for this particular engineering approach.
Begin
2021-01-27
WI
JT005211
Planned document number
prCEN/CLC/TR 17603-40-02
