Andreas Koch

E-Mail:	andreas.c.koch@tum.de
Phone:
Room:	9377.01.109
Address:	Airbus Defence and Space GmbH Andreas Koch Abteilung Telecom Processing Germany (TSEDH-TL2) Willy-Messerschmitt-Straße 9 82024 Taufkirchen

Research Interests

• Deep Learning
• Streaming Anomaly Detection on Satellite Telemetry
• Satellite Payload Data Processing
• AI hardware acceleration on FPGA

Projects

• Machine Learning Application Benchmark (MLAB): A European Space Agency (ESA)-funded activity by Airbus Defence and Space GmbH, Technical University of Munich and Orbital Oracle Technologies with the aim of creating a machine learning benchmark suitable for the space domain. It consists of multiple applications spanning anomaly detection, radio modulation classification, object detection for optical and multispectral images in order to detect ships, airplanes and wildfires, and land use / land cover classification. The benchmark combines aspects from both hardware and machine learning by offering different submission modes, allowing the comparison of either the machine learning model or the hardware platform.

• Anomaly Detection Anomaly Prognosis (ADAP): This was a project led by Airbus Defence GmbH in cooperation with Frauenhofer Institute for Integrated Circuits (IIS) in order to develop a machine learning (ML) based anomaly detection system for spacecraft telemetry on space-grade hardware. It included the compilation of a new labeled spacecraft anomaly detection dataset, the development of ML models, the implementation of an on-board computer with the capability of supervising ML applications and finally the accelerated deployment of ML models on space-grade hardware.

• Machine Learning for Telecommunication Satellites (MaLeTeSa): This Airbus Defence and Space project aims to develop a custom machine learning module for satellite communication processors and is funded by Deutsches Zentrum für Luft- und Raumfahrt (DLR). To this end, various ML approaches are researched and applications include anomaly detection in telemetry data, identification of different users of the radio frequency spectrum and optimization of communication networks.

Publications

Other Publications

1. Hagerer, G., Szabo, D., Koch, A., Dominguez, M. L. R., Widmer, C., Wich, M., Danner, H., & Groh, G. (2021). End-to-End Annotator Bias Approximation on Crowdsourced Single-Label Sentiment Analysis. In M. Abbas & A. A. Freihat (Eds.), Proceedings of the 4th International Conference on Natural Language and Speech Processing (ICNLSP 2021) (pp. 1–10). Association for Computational Linguistics. https://aclanthology.org/2021.icnlsp-1.1 [Online]