Coordinated beam management in massive MIMO systems for beyond 5G

Directional transmission in practical massive multiple-input-multiple-output (MIMO) systems allows to focus the signal energy on a particular area where the user is located. This approach not only brings the advantage of time and frequency reuse of the resources in the cell—and thus increasing the network capacity—but also improves the signal coverage. Beamforming is also particularly beneficial for high carrier frequencies, e.g. mmWave bands, where the free-space pathloss of the signal is high.

In electronically controlled beamforming approaches, e.g. butler matrix, a set of orthogonal beams known as code-book steer the beam into different mainlobe directions. Determining an efficient and reliable approach on how to select the best beam(s) from a given codebook is under investigation. The state-of-the-art beam alignment techniques such as exhaustive or tree search—as in IEEE 802.11ad—use a fixed length training sequence that optimally can be designed for a particular signal to noise ratio (SNR). Since the SNR value may vary depending on the user’s mobility or various conditions, an adaptive variable length training sequence would be of interest.

The tasks for this research work involves studying the above mentioned approaches and further investigating theoretical techniques for diversity gain in multi-cell scenarios. In a coordinated beam management scenario, multiple connections to different access points provide further diversity which is particularly of interest for ultra reliable communication.

Application of machine learning approaches, e.g. deep neural networks and their corresponding complexity comparisons with respect to the classical approaches are also welcome in this research.

Subtasks:

- literature research on beam selection techniques and stochastic channel models
- Derivation of novel algorithms for beam prediction in multi-cell scenarios.
- Derivation of novel channel equalization techniques.
- Simulation of the algorithms on Matlab or Python.
- Based on simulation results, possible implementations on hardware.
- Documentation and conclusion.

[1] W. Rave and M. K. Marandi, “The elimination game or: Beam selection based on m-ary sequential competition elimination,” in WSA 2019; 23rd International ITG Workshop on Smart Antennas, April 2019, pp. 1–8.

[2] M. K. Marandi, W. Rave, and G. Fettweis. "Beam Elimination Based on Sequentially Estimated a Posteriori Probabilities of Winning." ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2020.

Type of work: Research Project / Bachelor / Master

Requirements: Working independently, basic knowledge in linear Algebra, programming skills in Matlab or Python

Contact: Shahab Ehsan-Far

Illustration of beam elimination as discussed in [1], [2]. Here n is the sequence length and γ_m is the generalized log-likelihood ratio at beam m. As n increases, the slope of γ_m (n) for stronger beams is steeper than the slope of γ_(m') (n) of weak beams.