Marvin Rübenacke

Research

Error-correcting codes (channel codes) are the workhorse of modern communication systems, enabling reliable transmissions close to the theoretical limits. Emerging applications like industry automation, autonomous driving, remote surgery and many more require ultra reliable communication with low latency.

My research focusses on channel codes and decoding algorithms allowing low latency and high reliabality. For this reason, I examine codes with rich symmetries, whose decoding can be highly parallelized. Some key concepts I consider are:

• Polar codes and polar-like codes
• Low-density parity check (LDPC) codes
• Low latency, low complexity decoding algorithms, such as automorphism ensemble decoding (AED) and iterative decoding
• Optimization and Machine Learning

For more details, please have a look at my publications below.

Student Projects / Theses

I offer usually many bachelor, research or master thesis topics along the lines of

• Topics from my current research, such as code design, decoder design, decoder implementation, system design ...
• Webdemo conception and implementation
• Demonstrators

If you are a student interested in these or similar topics, feel free to contact me.

Teaching

I also enjoy building webdemos and interactive programs to make communications more approachable and exploring its concepts playful.

The following interactive program demonstrates the concept of geometric shaping. For different arrangements ("constellations") of the transmit symbols in the I/Q-plane, one can communicate different amount of information per symbol through an AWGN channel. Drag the constellation points with your mouse and see how close you can get to Shannon's capacity curve (hint: not only the positions, but also the labeling, i.e. the corresponding bit patterns, matter)!

The mutual information curve is computed via Monte Carlo simulation of thousands of bits in your browser in real time. Implemented in p5js.