Abstract:
Massive multiple-input multiple-output (MIMO) systems that employ a large number
of antennas at both receivers and transmitters have been widely considered for adoption
in next generation (5G) wireless networks. The deployment of massive MIMO promises
to enhance the received signal power for communications over millimeter-wave (mmWave)
spectrum, which in turn increases the throughput and system efficiency. Notwithstanding
the advantages of massive MIMO, several major technical challenges arise, which include
the difficulty and complexity in hardware implementation, precoder design and channel
estimation. In this thesis, we mainly focus on strategies that address the training overhead
issue for mmWave massive MIMO channel estimation. By utilizing the sparsity feature in
the angular domain of mmWave channels, we propose a gridless compressive sensing (CS)
technique based on atomic norm minimization (ANM). Particularly for massive MIMO
systems involving two-dimensional angle estimation, we develop a decoupled ANM (DANM)
approach that offers high-accuracy channel estimation at low complexity and little
training overhead. The proposed D-ANM approach is applied to mmWave massive MIMO
systems with uniform rectangular array employed at base station and extended to the multiuser
case. Investigation on the use of D-ANM for channel estimation in wideband mmWave
SIMO-OFDM systems is also carried out to cope with frequency-selective channel fading.
