This content is not published

Machine learning platform of spacecraft operational datasets

Tue, 12/05/2023 - 02:00
Start Date: 
2022
Programme: 
GSTP
End Date: 
2023
Programme Reference: 
GT1O-313GD
Country: 
Germany
Prime Contractor: 
TELESPAZIO VEGA DEUTSCHLAND GMBH
Status: 
Closed
Objectives: 
Create a production ready, user-friendly platform that is easily accessible to the user community for exploration and processing of spacecraft operations telemetry data using machine learning techniques and models.
Description: 
;;
Over the last decade the systems and tools available to the European operational community to perform long term storage, offline data analysis and processing have been extended and improved.
;
All these systems and tools have all been designed and developed mainly to serve the operations community internally and are therefore not suitable to be exposed as they are. Further, in case distribution of the data is restricted, the user community shall still be able to run their data analytics operations (including machine learning) on the complete dataset, and retrieve and share only the results. Therefore a major challenge for the activity is maintaining the necessary level of restriction and security, whilst allowing full exploitation of our datasets, which may require significant rework or extensions.
;
One of the common problems faced when trying to develop platforms to support data processing activities of operational data on demand is the definition of relevant use cases, establishment of boundaries acceptable to the users, and narrowing the selection of supported techniques and technologies. A core challenge of this activity is therefore to find a good representation of the community and capture their desires/requirements.
;
As a result of encouraging and supporting the sharing of algorithms, models and results within the community, industry can exploit available data sets to gain insights into space system performance, leading to simulation of improvements in design of space systems and mission operations.
;
The systems to be tested would apply to large distributed developments. The idea is to reuse automated test sequences and test drivers developed by different teams to test the integrated system.
;
The activity will consist of the following tasks:
    • Identify user needs scope and requirements
    • Definition of representative use cases from the user community
    • Analysis of the existing systems, architecture and classification restrictions to determine feasibility
    • Assess re-use and extension of existing systems and facilities for hosting and for data access.
    • Design of target architecture including: data repositories, security aspects, machine learning algorithms, framework environment, libraries and client interfaces. ;
    • Identification of the key underlying technologies and products that will support implementation of the design.
    • Detailed design of the architecture, including the deployment concept, redundancy concept and extensions to existing tools and systems.
    • Implementation of the design in an agile approach, from the core services outwards towards the user communities.
    • Implementation of a prototype including all the elements listed in the architecture. ;
    • Implementation of the test framework procedures (not spacecraft procedures) and guidelines to develop the automated test sequences so that they can be reused and exchanged (a set of common rules). ;
    • Demonstration of the concept in a simulation environment for the test cases selected.
    • Implementation of a robust CI/CD pipeline to enable rapid deployment and validation of software fixes
    • Application of solution to selected communities and missions.
;
;
Application Domain: 
Generic Technologies
Technology Domain: 
9 - Mission Operation and Ground Data Systems
Competence Domain: 
8-Ground Systems and Mission Operations
Keywords: 
47-OPS-INN
Initial TRL: 
TRL 4
Target TRL: 
TRL 7
Achieved TRL: 
TRL 6
Public Document: