• Themenbeschreibung: The LIKE chair is working on new channel access schemes for 5G New Radio (NR), which offers a wide range of use cases such as enhanced Mobile Broadband (eMBB), Ultra-Reliable Low-Latency Communication (URLLC), and massive Machine Type Communication (mMTC). These use cases impose new challenges on 5G such as massive connectivity, high spectral efficiency, reliability, and ultra-low latency. The conventional Orthogonal Multiple Access (OMA) schemes like TDMA, FDMA, and CDMA fail to meet the challenging requirements of 5G due to the scarcity of resources. Non-Orthogonal Multiple Access (NOMA) schemes on the other hand, allow the users to share the resources, yielding a more efficient utilization. However, the shared access results in collisions between the transmitted data of different users. As a result, more sophisticated Multiuser Detection (MUD) schemes are required in order to the resolve the collisions.
    The exhaustive complexity of the optimum Maximum A Posteriori (MAP) receiver, that grows exponentially in the number of users and the modulation size, renders it unfeasible. Expectation Propagation Algorithm (EPA) is an iterative message-passing algorithm that offers a reasonable alternative to the optimum MAP decoder. Expectation Propagation (EP) is an inference technique that was initially proposed for machine learning. The idea behind EP is to project a sophisticated posterior distribution onto a tractable family of distributions like the Gaussian distribution. This means that the bottleneck of the decoding complexity of multiuser detection, which resides in calculating the posterior distributions, translates to analyzing some Gaussian distributions. The iterative decoding algorithm is carried out by transmitting messages back and forth on what is so called the Tanner Graph. These messages simply correspond to the updated mean and variance of the approximated Gaussian at each decoding iteration. The scope of this project is to implement the expectation propagation algorithm using C++, and compare its performance and complexity to the state-of-the art optimum receivers.
  • Conducting a detailed research on the state-of-the-art receivers of NOMA
  • - Implementing the Expectation Propagation Algorithm using C++
  • - Analyzing the decoding performance as a function of the number of decoding iterations of the Expectation Propagation Algorithm
  • - Optimizing the tradeoff between the decoding complexity and the accuracy of the approximation (projecting the exact posterior distribution onto Gaussian)
  • - Comparing the performance of the Expectation Propagation Algorithm to that of the optimum MAP decoder
  • Themengebiete:
  • Voraussetzungen:
  • Betreuer: Sally Nafie
  • Hochschullehrer:




    Anfragen zu Themen und der Betreuung einer Forschungspraktika können sie auch unabhängig von den ausgeschriebenen Themen direkt an die Ansprechpartner der jeweiligen Forschungsschwerpunkte schicken.


    Dr.-Ing. Jörg Robert

    Am Wolfsmantel 33
    91058 Erlangen - Tennenlohe


    Dipl.-Ing. Jürgen Frickel

    Am Wolfsmantel 33 / 3.OG
    91058 Erlangen - Tennenlohe

    Navigation und Ortsbestimmung / Autonome Robotik

    Adam Kalisz

    Am Wolfsmantel 33 / 3. OG
    91058 Erlangen - Tennenlohe

    Navigation und Ortsbestimmung / Funkortung

    Sebastian Kram

    Am Wolfsmantel 33 / 3. OG
    91058 Erlangen - Tennenlohe

    Navigation und Ortsbestimmung / GPS und Allg.

    Prof. Dr.-Ing. Jörn Thielecke

    Am Wolfsmantel 33 - 3.OG
    91058 Erlangen - Tennenlohe