Resource Optimization in Distributed Massive MIMO under Signaling and Hardware Limitations
Thesis event information
Date and time of the thesis defence
Place of the thesis defence
L10, Linnanmaa
Topic of the dissertation
Resource Optimization in Distributed Massive MIMO under Signaling and Hardware Limitations
Doctoral candidate
Master of Technology Bikshapathi Gouda
Faculty and unit
University of Oulu Graduate School, Faculty of Information Technology and Electrical Engineering, CWC-RT
Subject of study
Communications Engineering
Opponent
Professor Luca Sanguinetti, University of Pisa, Italy
Custos
Professor Antti Tölli, University of Oulu
Resource Optimization in Distributed Massive MIMO under Signaling and Hardware Limitations
Distributed massive multiple-input multiple-output (MIMO) systems consist of a number of access points (APs) or remote radio heads (RRHs) distributed across an area, connected to the central units via fronthaul links for the data and channel state information (CSI) exchange, coherently serving user equipments (UEs) in the network. Ideally, this setup eliminates the inter-cell interference and provides uniform service to all the UEs.
In the first part of the thesis, a fully distributed beamforming design framework is introduced to address the limitations of fronthaul signaling associated with the CSI exchange for the cooperative beamforming design. For cooperative precoding design at the APs in the downlink (DL), a novel over-the-air (OTA) signaling resource is introduced to obtain the same cross-term CSI information traditionally exchanged via the fronthaul links. Specifically, by utilizing a bidirectional training procedure with the proposed OTA signaling resource, network-wide precoders can be computed locally at each AP. Furthermore, the framework is expanded to uplink (UL) to include cooperative combining designs at the APs.
In the second part of the thesis, a bi-directional training procedure is introduced to design both DL and UL beamformers together, using both centralized and distributed methods. This approach improves the effective data rate by reducing the overhead associated with separate training for the DL and UL beamformers. Furthermore, the training resources for DL and UL distributed beamforming design are minimized by devising a single multicast beamformer to serve pairs of UEs, active either in DL or UL.
In the third part of the thesis, the UE transmit powers in the UL are optimized, considering low-cost, power-efficient RF hardware at the RRHs. Due to the use of 1-bit analog-to-digital converters (ADCs), non-Gaussian quantization distortion is introduced, rendering a nonmonotonic signal-to-interference-plus-noise-and-distortion ratio (SINDR) behavior with respect to the UE transmit power. The SINDR analysis demonstrates that adding appropriate dithering at the RRHs results in unimodal SINDR behavior. Consequently, various methods are proposed to jointly optimize the UE transmit powers and the RRH dithering levels to enhance the UL system performance.
The fully distributed beamforming designs proposed in this thesis outperform the noncooperative approaches, and offer scalability that supports practical implementation, even in large networks. Additionally, RRHs equipped with 1-bit ADCs can significantly benefit from joint reception across multiple RRHs in a range of scenarios.
In the first part of the thesis, a fully distributed beamforming design framework is introduced to address the limitations of fronthaul signaling associated with the CSI exchange for the cooperative beamforming design. For cooperative precoding design at the APs in the downlink (DL), a novel over-the-air (OTA) signaling resource is introduced to obtain the same cross-term CSI information traditionally exchanged via the fronthaul links. Specifically, by utilizing a bidirectional training procedure with the proposed OTA signaling resource, network-wide precoders can be computed locally at each AP. Furthermore, the framework is expanded to uplink (UL) to include cooperative combining designs at the APs.
In the second part of the thesis, a bi-directional training procedure is introduced to design both DL and UL beamformers together, using both centralized and distributed methods. This approach improves the effective data rate by reducing the overhead associated with separate training for the DL and UL beamformers. Furthermore, the training resources for DL and UL distributed beamforming design are minimized by devising a single multicast beamformer to serve pairs of UEs, active either in DL or UL.
In the third part of the thesis, the UE transmit powers in the UL are optimized, considering low-cost, power-efficient RF hardware at the RRHs. Due to the use of 1-bit analog-to-digital converters (ADCs), non-Gaussian quantization distortion is introduced, rendering a nonmonotonic signal-to-interference-plus-noise-and-distortion ratio (SINDR) behavior with respect to the UE transmit power. The SINDR analysis demonstrates that adding appropriate dithering at the RRHs results in unimodal SINDR behavior. Consequently, various methods are proposed to jointly optimize the UE transmit powers and the RRH dithering levels to enhance the UL system performance.
The fully distributed beamforming designs proposed in this thesis outperform the noncooperative approaches, and offer scalability that supports practical implementation, even in large networks. Additionally, RRHs equipped with 1-bit ADCs can significantly benefit from joint reception across multiple RRHs in a range of scenarios.
Last updated: 9.12.2024