Laufende Forschungspraktika


Im Forschungspraktikum wird die Praxis wissenschaftlichen Arbeitens in der Forschung am Lehrstuhl für Digitale Übertragung vermittelt. Der Schwerpunkt der Forschungsarbeiten kann experimentellen, theoretischen oder auch konstruktiven Charakter haben. Kombinationen aus unterschiedlichen Schwerpunkten sind auch möglich.

Ein Forschungspraktikum ist in vielen Masterstudiengängen möglich. Viele unserer angebotenen Masterarbeiten können auch als Forschungspraktikum durchgeführt werden. Bitte wenden Sie sich bei Interesse an die jeweiligen Mitarbeiter. Es existieren oft auch Arbeiten, die noch nicht ausgeschrieben wurden.

Projects:

  • Achievable Rate Region of the Bidirectional Full-Duplex Channel (ongoing)

    Giuseppe Rindone

The proposal can be found here.

  • Drone-Based Communication Using FSO Backbone (ongoing)

Hedieh Ajam, since 01.07.2016

The proposal can be found here.

  • Interference Alignment for SC-FDMA Systems with Widely Linear Filtering (ongoing)

Mohamed Soliman, since 11.2016

In expectation of a growing interference level in cellular systems in the near future, elaborate interference management techniques are necessary in addition to the traditional interference avoidance approaches employed so far. A first step has been done when Coordinated Multi-Point (CoMP) transmission was introduced in the 4G LTE-A standard, allowing an exchange of information between base stations from neighbouring cells. For the network of the next generation, CoMP will additionally embrace most likely centralized processing and distributed cooperation. This new architecture gives way to more advanced cooperation techniques such as interference alignment. Interference alignment is a technique that recently has attracted attention due to its capability to theoretically increase the sum rate of a network without any bound. This is realized by jointly designing precoding filters for the transmitters such that the interference falls into a reserved subspace at the receiver while leaving the remaining subspace interference-free for the desired user. Although theory shows that a complete and perfect alignment is possible, under real conditions residual interference still will be present due to a low signal-to-noise ratio and a limited number of signaling dimensions. Task of the research internship is to investigate widely linear (WL) filtering at the transmitter and the receiver side to overcome the issue associated with limited number of signaling dimensions. For WL filtering the imaginary and the real part of the filter input are processed seperately and subsequently combined linearly. When additionally applying real-valued transmit symbols, a WL system model results wich has doubled dimensions. This translates into more degrees of freedom for the filter design. As the joint optimization over precoding, receive matrices of the users implies an exhaustive search, a suboptimal approach will be examined. The transceiver design is conducted for the 4G formats single-carrier frequencydivision multiple access (SC-FDMA).

  • Deep Learning for Equalization and Cochannel Interference Cancellation (ongoing)

Ilse Sofia Ramirez, since 08.02.2018

Recently, machine learning techniques have received significant attention in signal processing. For example, it has been shown that deep learning schemes offer an excellent performance in some applications in audio and multimedia signal processing. In this research internship, the potential of deep learning will be investigated for some selected  applications in signal processing for communications. In particular, the equalization of signals received over highly frequency-selective channels and the recovery of signals from observations impaired by severe cochannel interference, respectively, will be considered. After a review of deep learning approaches, some suitable schemes will be selected and implemented. The influence of training on the performance of the selected schemes will be analyzed, and a comparison with conventional schemes with respect to performance and complexity will be conducted.