Area of research:
Have you (almost) completed your degree studies? Are you inspired by space exploration and keen to pursue a career in this exciting field? If so, perhaps you should take a closer look at the German Trainee Programme. Organised by DLR, it offers you the chance to work shoulder to shoulder with experts from the 22 member states of ESA – keeping your finger on the pulse of Europe‘s space programmes. Over a period up to 24 months, you will actively contribute to the latest research and/or technology projects. This is complemented by a generous scholarship. What better way to launch your career in international space business? The next GTP commences on 1st February 2020.
This is your opportunity to join the team at ESTEC – Noordwijk, Netherlands
Deployment of Machine Learning in On-Board Software Systems for Payload Processing (GTP-2020-TEC-EDP)
The TEC-EDP section carries out work in the field of payload data systems, high-performance payload processing systems, image processing and compression, on-board processing algorithms of payload data, as well as high-speed interfaces and networks. As part of ESA’s Directorate of Technical and Quality Management, the section provides technical support in the areas of competence to all programme directorates in order to support their missions and developments. In addition, the section is responsible for initiation and execution of technology development activities that may lay the foundation for future missions.
Currently, research and development work is carried out to enable complex on-board processing through the development of high-performance processing systems based on multi- and manycore processors, FPGAs, etc. Activities in the fields of: benchmarking of on-board processing applications on low-power GPUs; development of high-performance fault-tolerant payload processing systems based on COTS- processors; and studying future information extraction algorithm for advanced on-board processing are currently on-going. The development of new high-performance processing systems and adaption of payload processing algorithms for embedded systems, will be a crucial part in enabling complex processing, such as machine learning, on-board for payload applications.
You will be placed in the on-board payload data handling and processing laboratory, which is a fully equipped lab for high-performance payload systems.
Both entities, DLR-SC and ESA TEC-ED, research in the area of machine learning in embedded systems targeting the space domain. You will work in this field and bridge the two entities, to enable knowledge and idea exchange. This will further condense ESA’s and DLR’s perspective on AI in space, allowing us to be more effective in shaping the future in that emerging field in Europe.
New proposed mission concepts could enable the spacecraft to make decisions for the prioritization of science data on the data downlink is just one proposal. While this represent only the tip of the iceberg for AI technology in the domain, DLR and ESA expects a strongly growing demand in the upcoming near future for the concrete application of AI on spacecraft and wants to be prepared.
Currently, there is a need to understand how machine learning algorithms can be deployed in current and future on-board processing systems, as the embedded systems put specific constraints on the potentially usable machine learning algorithms.
You will study how the deployment process of machine learning application for on-board payload processing affects the constraints and requirements of the on-board systems.
In particular, throughout the project duration, you will develop in the field, in particular:
understanding the pros and cons of various Artificial Intelligence (AI) technologies; and studying the current work of the use of AI technologies in on-board systems identification of potential machine learning technologies, combined with conventional algorithms to meet the requirements for space systems practical knowledge of the computational limitations of on-board systems for machine learning tasks, and needed adaption and optimization of neural networks of embedded systems benchmarking of machine learning models performance after adaption of target systems understanding the end-to-end development flow for advanced on-board processing algorithm development and deployment; including the interfaces and work distribution among the ground andspace segment of the satellite system considering on-board payload image processing as a use-case, developing demonstrator software to reveal the capabilities of machine learning algorithms on a spacecraft evaluating validation, verification, and qualification strategies for deep learning algorithms for space applications publication of the results at international conferences