E-Mail: | michael.petry@tum.de |
Phone: | 4285 |
Room: | 9377.01.113 |
Address: |
Airbus Defence and Space GmbH Michael Petry Abteilung Telecom Processing Germany (TSTCG-TL2) Willy-Messerschmitt-Straße 9 82024 Taufkirchen |
Machine Learning for Telecommunication Satellites (MaLeTeSa): This DLR-sponsored project aims to explore how artificial intelligence can benefit telecommunication satellites. It comprises two components. The first component focuses on exploring and developing machine-learning-based software and algorithms in the fields of wireless communication signal processing, anomaly detection, and network orchestration, while the other one aims at developing a space-grade System-on-Chip-based AI-Processor on which the algorithms shall be deployed on-board the satellite. The main focus of my doctorate is both developing AI-based RF algorithms and implementing them on the hardware platform. The result is a hybrid processing pipeline that comprises both neural networks and classic DSP components, which is then realized on the heterogeneous hardware platform. By distributing the processing steps between FPGA, AI-Engines (ASIC-like vector processors), and CPU, and ensuring optimal data transfer between those components, an efficient implementation is achieved.
Artificial Intelligence for Anti-Jamming (AJAI): In this ESA-sponsored project we explore how artificial intelligence-based techniques can be applied to achieving a robust satellite communication link that is particularly resilient to external signal jamming attacks. By focusing on partial band, multi-tone, follower, and sweeper jammer scenarios, we design a waveform comprising an AI-assisted feedback loop and implement it on the space-grade AI-processor hardware which is developed within the MaLeTeSa project (see above).
5G Autosat KI: In this DLR-sponsored project we implement a 3GPP standard-compliant 5G basestation (gNodeB) cellular stack on a System-on-Chip (FPGA, AI-Engines, CPU)-based space-grade hardware platform and augment it with AI processing capabilities. After completing the AI-related steps of this project (designing, training, and deployment of a proof-of-concept traffic prediction algorithm based on 5G user meta-data), I am focusing on extending the hardware setup with a high-performance RF interface to improve maximum communication speeds.