Intelligent Reflecting Surfaces and Advanced Multiple Access Techniques for Multi-AntennaWireless Communication Systems
Sousa de Sena, Arthur (2022-10-24)
Väitöskirja
Sousa de Sena, Arthur
24.10.2022
Lappeenranta-Lahti University of Technology LUT
Acta Universitatis Lappeenrantaensis
School of Energy Systems
School of Energy Systems, Sähkötekniikka
Kaikki oikeudet pidätetään.
Julkaisun pysyvä osoite on
https://urn.fi/URN:ISBN:978-952-335-860-7
https://urn.fi/URN:ISBN:978-952-335-860-7
Tiivistelmä
Multiple-input multiple-output (MIMO) is an indispensable technology for deploying the pervasive connectivity sought for fifth-generation (5G) and beyond communication systems. By relying on a large number of antennas, massive MIMO schemes can implement space division multiple access (SDMA) to serve spatially separated users with a single frequency–time resource block, thus, leading to incredible spectral and latency enhancements. Nevertheless, there are certain communication scenarios, such as ultradense deployments or environments with users sharing overlapping angular positions, where spatial multiplexing becomes unrealizable through SDMA alone. These challenging scenarios motivate the exploitation of different domains and technologies. In particular, power-domain non-orthogonal multiple access (NOMA) and rate-splitting multiple access (RSMA) have appeared as strong candidates for extending the capabilities of MIMO systems and enabling resource-efficient simultaneous transmissions even to overlapping users. In parallel development, a disruptive concept of an intelligent reflecting surface (IRS) has arisen as a method to manipulate electromagnetic propagation through reconfigurable, low-power, subwavelength reflecting elements. As their main feature, the properties of IRSs can be dynamically tuned and harnessed to attack harsh phenomena of wireless channels and accomplish diverse objectives, enabling communication environments with optimized signal radiation. Driven by the promising capabilities of the above-mentioned technologies, this doctoral dissertation focuses on studying and developing novel transmission schemes based on the synergy between NOMA, RSMA, and IRSs, and their application to next-generation multiuser MIMO communication networks.
This research work starts by investigating practical issues of imperfect successive interference cancellation (SIC) on a downlink multicluster massive MIMO-NOMA network. Through an in-depth theoretical analysis, exact closed-form expressions are derived for the outage probability and ergodic rates observed by each user. By exploiting the Karush– Kuhn–Tucker conditions, efficient dynamic power allocation strategies are also implemented for improving rate fairness within each cluster in the network. Motivated by the performance limitations identified in our seminal investigations, our research is continued on the MIMO-NOMA topic by exploiting the powerful capabilities of IRSs to tackle the interference issues of SIC. To broaden our optimization opportunities, a novel disruptive dual-polarized IRS is proposed to harness the additional degree of freedom offered by the polarization domain. By manipulating wave polarization with these promising IRSs through interior-point and conditional gradient methods, advanced dual-polarized transmission strategies are implemented, which can effectively mitigate SIC-related problems and remarkably improve the data rates of all users, both in the downlink and uplink of dual-polarized MIMO-NOMA networks. Among our contributions for the downlink, besides optimizing the IRS reflecting elements, a closed-form expression is derived for the ergodic rates considering large IRSs, whereas for the uplink, also a low-complexity alternate power allocation policy is proposed for balancing uplink data rates.
Next, motivated by the impressive broader region of achievable rates possible with RSMA, this research is advanced and the advantages of the amalgamation between IRSs and MIMO-RSMA are investigated, showing that SIC issues can also degrade the performance of RSMA-based schemes. To solve the limitations introduced with SIC once and for all, in our last results, a novel high-performance dual-polarized massive MIMORSMA scheme is proposed that does not require SIC, thereby eliminating all associated problems. As a practical tool for assisting the design of the proposed system, a deep neural network (DNN) model is implemented for predicting the ergodic sum-rates observed in the network with high accuracy. Last, an intelligent DNN-aided adaptive power allocation strategy is developed, which maximizes the sum-rate of the dual-polarized MIMO-RSMA even under high levels of cross-polar interference and imperfect channel state information. Our contributions and all novel transmission schemes proposed in this doctoral dissertation are supported with extensive simulation results and fair performance comparisons with state-of-the-art baseline communication systems.
This research work starts by investigating practical issues of imperfect successive interference cancellation (SIC) on a downlink multicluster massive MIMO-NOMA network. Through an in-depth theoretical analysis, exact closed-form expressions are derived for the outage probability and ergodic rates observed by each user. By exploiting the Karush– Kuhn–Tucker conditions, efficient dynamic power allocation strategies are also implemented for improving rate fairness within each cluster in the network. Motivated by the performance limitations identified in our seminal investigations, our research is continued on the MIMO-NOMA topic by exploiting the powerful capabilities of IRSs to tackle the interference issues of SIC. To broaden our optimization opportunities, a novel disruptive dual-polarized IRS is proposed to harness the additional degree of freedom offered by the polarization domain. By manipulating wave polarization with these promising IRSs through interior-point and conditional gradient methods, advanced dual-polarized transmission strategies are implemented, which can effectively mitigate SIC-related problems and remarkably improve the data rates of all users, both in the downlink and uplink of dual-polarized MIMO-NOMA networks. Among our contributions for the downlink, besides optimizing the IRS reflecting elements, a closed-form expression is derived for the ergodic rates considering large IRSs, whereas for the uplink, also a low-complexity alternate power allocation policy is proposed for balancing uplink data rates.
Next, motivated by the impressive broader region of achievable rates possible with RSMA, this research is advanced and the advantages of the amalgamation between IRSs and MIMO-RSMA are investigated, showing that SIC issues can also degrade the performance of RSMA-based schemes. To solve the limitations introduced with SIC once and for all, in our last results, a novel high-performance dual-polarized massive MIMORSMA scheme is proposed that does not require SIC, thereby eliminating all associated problems. As a practical tool for assisting the design of the proposed system, a deep neural network (DNN) model is implemented for predicting the ergodic sum-rates observed in the network with high accuracy. Last, an intelligent DNN-aided adaptive power allocation strategy is developed, which maximizes the sum-rate of the dual-polarized MIMO-RSMA even under high levels of cross-polar interference and imperfect channel state information. Our contributions and all novel transmission schemes proposed in this doctoral dissertation are supported with extensive simulation results and fair performance comparisons with state-of-the-art baseline communication systems.
Kokoelmat
- Väitöskirjat [1038]