Research On Multi-Domain Wireless Resource Scheduling And Optimizations For Millimeter-Wave MIMO-NOMA Systems

Driven by the overwhelming increase of wireless devices and the insatiable appetite for ubiquitous smart wireless connections,next-generation wireless communications are seeking for revolutionary technologies to enable ultra-high-speed communications,massive access,and personalized communication experience.To realize this aspiration,millimeter-wave(mmWave)multiple-input multiple-output and non-orthogonal multiple access(MIMO-NOMA)has emerged as a promising technique to overcome the radio spectrum depletion and the inefficiencies of traditional multiplexing technologies.While existing researches have laid solid contributions,mmWave MIMO-NOMA systems still suffer from several challenges,including but not limited to:(1)Conventional channelbased mmWave MIMO-NOMA methods suffer from severe communication instability and uncontrollable quality of service(QoS)performance in the high-frequency wireless environment.(2)Existing unlicensed-band coexistence approaches can only passively avoid interference among heterogeneous mmWave networks,failing to efficiently mitigate the mutual interference between densely connected mmWave users.(3)Conventional MIMO-NOMA approaches cannot realize efficient radio resource scheduling in a complex generalized mmWave communication environment,which may experience different regimes of channel correlations and system loadings.Against this background,this dissertation attempts to investigate spectral-and energy-efficient mmWave MIMO-NOMA radio resource scheduling and optimization methods under the dynamic,heterogeneous,and complex wireless communication environment.The main research contents include:(1)A novel dynamic reconfigurable mmWave MIMO-NOMA access method is proposed to address the communication instability under dynamic wireless environment.Specifically,the unreliable and uncontrollable dynamic wireless environment is smartly reconfigured via the coordination among multiple mmWave access points(APs)and reconfigurable intelligent surfaces(RISs),thus enabling on-demand and quality-of-service(QoS)-driven mmWave MIMO-NOMA access for different users.The dynamic reflecting element selection,coordinated discrete phase-shift control,and power allocation are jointly optimized to maximize the system energy efficiency,while ensuring users’ diverse data rate and reliability requirements.To realize distributed coordination among the partially observable agents,a novel graph-embedded multi-agent deep reinforcement learning algorithm is proposed,which embeds agents’ local observations into lowdimensional messages to enhance multi-agent coordination.It is theoretically demonstrated that the proposed algorithm can improve the generalization ability as well as converging to a locally optimal policy.Numerical results reveal the efficiency of the proposed method in terms of both system energy efficiency and QoS guarantees.(2)A novel proactive coordination and coexistence method is proposed for heterogeneous mmWave system over the 60 GHz unlicensed band.Due to the openness of the unlicensed spectrum,the 60-GHz unlicensed-band mmWave system will be a heterogeneous system,which contains both cellular and Wireless Gigabit(WiGig)networks complying with utterly different protocols and standards.Thus,direct co-channel interference coordination becomes unrealizable.Different from existing over-protected directional Listen-Before-Talk(LBT)methods that can only avoid interference passively and non-coordinately,the proposed method enables cellular APs to proactively mitigate interference and coordinate MIMO-NOMA resource scheduling by observing the co-channel WiGig signals based on IEEE 802.11 ay protocol.Hence,efficient spectral utilization can be achieved under dense user connections.To maximize the system spectral efficiency(SE)while ensuring WiGig network performance and cellular users’ delay requirements,a penalty dual decomposition and convex-concave procedure(PDDCCCP)algorithm is developed to jointly optimize the user clustering,hybrid beam coordination,and power control.Numerical results validate that the proposed method can significantly enhance SE and reduce delay under dense user connections,while achieving comparable interference suppression capability with conventional directional LBT.(3)A novel environment-adaptive MIMO-NOMA radio resource optimization method is proposed,which can adapt to the complex generalized communication environment with various regimes of channel correlations and/or system loadings.By eliminating the conventional limitation of sequentially performing successive interference cancellation(SIC)within pre-defined cluster(s),the proposed method can unify and generalize the existing environment-specific MIMO-NOMA methods,thus enabling highly flexible multi-domain multiplexing and a novel environment-adaptive mmWave MIMO-NOMA communication paradigm.To maximize the system sum rate for a single-AP network,the hybrid beamforming and SIC operations are jointly optimized via an alternatingdirection multiplier method.Considering the more realistic multi-AP network,a lowcomplexity distributed automated-learning graph neural network(Auto GNN)algorithm is further proposed to jointly optimize multi-AP coordinated hybrid beamforming and SIC operations.The proposed Auto GNN can automatically self-optimize the graph neural network(GNN)structure during the GNN model training,thus reducing both computation and communication overheads whilst optimizing the system performance.Numerical results verify the effectiveness of the proposed environment-adaptive mmWave MIMO-NOMA method in diverse complex communication environment.The main research achievements and findings of this dissertation can contribute to methodological designs and theoretical analyses for radio resource scheduling and massive access toward next-generation wireless communications.