Study on Reinforcement Learning for Next-generation (6G) wireless communications


Next-generation wireless systems are expected to autonomously utilize technologies in a distributed fashion through potential to exploit machine learning techniques. The performance of deep learning models degrades with time in dynamic wireless environments due to the ageing effect [1]. The ultra-reliable low-latency communication (URLLC) requirement of interactive Industry 4.0 applications such as virtual reality acts as a challenge for application of supervised learning models. The results in [1] show that supervised and unsupervised learning models, in general, are susceptible to the degradation of performance due to ageing. Even though reinforcement learning techniques show promising results, the low latency requirement and the high computational requirement act as limitations for practical implementations.


Here at the Chair of Distributed Signal Processing (DSP) we have several master thesis offers to study reinforcement learning (RL) algorithms have shown to perform well under dynamic environments and perform efficiently. However, to cope with the low latency and high-reliability requirements of modern applications, these models must be optimized further. Our objective is to study RL techniques through which the training process can be optimized in order to make them applicable in a rate constrained communication system.


Among others these steps need to be done:

  • Prepare laboratory practicum with description of C-based and SDR-based exercises based on drafted examples
  • Prepare laboratory exercises based on the existing C-code examples of communication systems
  • Test and validate the implementation of online C-based laboratory framework (compiler/server/user)
  • Test and validate the implementation of online SDR laboratory framework (SDR/server/user)


Dr. H. Kremo



Depending on the actual task, the following is essential:

  • Good programming skills in C / C++ (object oriented programming) and Python (or other ML library)
  • Experience with traditional optimization and machine learning techniques
  • Good understanding of the physical layer processing for communications

In case of interest, please send me an email including the following:

  • Latest transcript of records
  • A brief description of your background and motivation
  • CV (if available)