Massive MIMO (M-MIMO), which consists of a large number of antennas at the base station (BS), is a promising technology to meet the high data rate and quality of service requirements of 5G wireless systems, while the hybrid beamforming architecture reduces the implementation costs with a low dimensional baseband processing unit combined with a massive phase array. In this project, we investigated how to reduce the BS power via 1) QoS-aware eigenmode selection hybrid beamforming and 2) machine learning based adaptive radio frequency (RF) chain switching. We focus on a downlink communication scenario where a hybrid beamforming (BS) transmits data to multiple single-antenna users.
QoS-aware Eigenmode Selection Hybrid Beamforming (published):
In this part, given the knowledge of channel state information of all users, the hybrid beamformers at the BS are designed to minimize the BS power consumption while the data rate needed to meet quality of service (QoS) requirement of each user is satisfied. Herein, the zero-forcing (ZF) beamforming is directly applied on the effective baseband channel, and the RF beamformer is generated by matching the beamforming matrix columns, selected from a preset discrete Fourier transform (DFT) basis codebook, with the eigenvectors of the aggregated propagation channel, which is termed as eigenmode selection. We also present a phase array batch-switching structure to realize the eigenmode selection beamforming economically. Simulations demonstrate that substantial transmission power can be saved with the proposed eigenmode selection beamforming compared to existing propagation path matching schemes, especially in rich-scattering channels, while satisfying given QoS (data rate) requirements.
Machine learning based adaptive RF chain switching (on-going):
In this part, of our interests is to shrink the power consumption of a hybrid beamforming MIMO BS in the multi-user downlink communication, via regulating the budgets of the transmission power and the RF chains circuit power in a mixed-timescale fashion. Given the instantaneous channel state information (CSI) and the user data rate requirements, we applied the proposed QoS-aware eigenmode selection hybrid beamforming to reduce transmission power. Considering the statistical CSI in a large timescale, we adapt the number of active RF chains to minimize the total average BS power, with the outage probability constraint satisfied. A partition-based machine learning framework, with the extracted features, called power coefficients, is designed to track the optimal RF chain configuration. Simulations demonstrate that substantial transmission power can be reduced with the EiS-HBF scheme, while the adaptive RF chain switching technique is also effective to track the optimal RF configuration to significantly cut down the total average BS power.