Thesis description

Wireless communication is one of the most dynamic areas which has seen great technological leaps starting from 2G all the way to today’s buzzword 5G. Many new use cases of communications are emerging to drive the 5G and beyond technologies such as massive Internet-of-Things (IoT), ultra-reliable low latency communications (URLLC) and ever evolving enhanced mobile broadband (EMBB). All these novel applications have challenging requirements such as serving high-density of users/devices simultaneously, increased rates per device and reduced latency for mission critical applications. Beam-forming is a powerful technique to boost the desired signal towards intended users while suppressing harmful interference and thereby helpful in addressing above challenges. However, traditional linear iterative algorithms for beamforming may have significant disadvantage in terms of robustness against channel state information error, computational complexity and/or latency when dealing with increasing number of antennas and/or cooperating base-stations in 5G networks. Therefore, there is a great interest towards machine learning (ML) based approaches for designing the downlink beamforming which can be trained efficiently (e.g. offline unsupervised), achieving better performance vs complexity trade-offs and have mechanisms for adapting to dynamic channel conditions and/or scenarios. In this respect, the goals of this project are as follows:

• Goal 1: Literature review of existing ML-approaches for 5G beam-forming
• Goal 2: Implementing the training and testing of chosen ML-approaches in Python environment
• Goal 3: Performance comparison of studied ML-approaches with traditional algorithms

Qualifications

• Master student in Electrical Engineering, Computer Science or equivalent.
• A solid theoretical background in machine learning
• Experience in modeling, MATLAB and Python.

Location: Kista, Stockholm

Preferred starting date: November, 2022

Extent: 1 student, 30hp.

Contact person: Ahmet Gokceoglu, ahmet.hasim.gokceoglu1@huawei.com