Beam alignment in multiuser millimeter wave communication systems
This thesis considers the multiuser beam alignment problem for MIMO millimeter wave (mmWave) communication systems. This problem is challenging because in mmWave systems beams are narrow and as such beamforming has a key role.
One way to approach such a problem is devising compressive sensing-based methods due to the sparse nature of mmWave channels. The sparsity results from the combination of a small number of clusters and a large number of transmit and receive antenna elements. In compressive sensing-based methods a high-dimensional vector is estimated using a lower-dimensional measurement vector. However, compressive sensing-based methods need highly time-accurate and high-rate control signals, bringing about large overhead.
Another way to approach the beam alignment problem is designing efficient beam search algorithms. In these methods, beamforming codebooks are used or designed for the base station and the user equipment (UE), and an efficient search of beamforming codebooks is performed to find the beams for both the base station and UE.
This thesis presents novel deterministic compressive sensing-based approaches for beam alignment. Firstly, a sparse and Kronecker-based sensing matrix, which is computationally efficient, is proposed. The proposed sensing matrix, namely matrix-bymatrix Kronecker product (MbMKP) sensing matrix, is constructed by computing the Kronecker product of two sensing matrices corresponding to the base station and UE. It is shown that the MbMKP sensing matrix satisfies the restricted isometry property, guaranteeing the reconstruction of the unknown sparse vector. It is also shown that the proposed approach outperforms random beamforming techniques in practical low signal-to-noise ratio (SNR) conditions.
Secondly, an algorithm is proposed to design a better deterministic sensing matrix, namely row-by-row Kronecker product (RbRKP) sensing matrix, that has advantages over the MbMKP sensing matrix. Not only does the RbRKP sensing matrix have flexibility in terms of the number of measurements but also it results in a larger number of distinct beamforming vectors. The implementation of compressive sensing-based approaches for beam alignment is also considered. Two different time scales for beamswitching are examined. In addition, different time scales for running the compressive sensing algorithm at UE are compared, taking into account their corresponding overhead and complexity. An overarching trial-based protocol that re-initializes the beam alignment procedure at particular times is proposed for implementing compressive sensing approaches.
Beam search approaches for multiuser mmWave systems that employ analog beamforming are also proposed. In the downlink, UEs perform a search of their beams independently. Then in the uplink, the beam directions of each UE are found by the base station simultaneously. Furthermore, a joint UE and beam acquisition approach in which the base station accepts strong UEs and rejects weak UEs during the beam search is proposed to expedite the beam alignment. This approach is designed based on posterior probability that a base station beam is weak for a UE, or based on comparing two hypotheses corresponding to a base station beam being weak and strong for a UE.