Research On Non-orthogonal Multiple Access In Enhanced Mobile Broadband And Massive Access Scenarios

In order to achieve 1000x improvement in mobile data rate and a 100x increase in device connection density,the fifth generation mobile communication system(5G)conducts technical research mainly from three aspects:ultradense heterogeneous networking,spectrum efficiency improvement and spectrum re-source exploitation.Non-orthogonal multiple access(NOMA),as a key technology for spectrum efficiency improvement,plays a significant role in enhancing 5G data rate and connection density.The enhanced mobile broadband scenario and massive access scenario are typical scenarios of 5G.In the above two scenarios,NOMA faces the challenge of reducing inter-user interference in the case of coordinated and uncoordinated access.The enhanced mobile broadband scenario aims at providing high data rate and satisfying differentiated require-ments of wireless services.User pairing and the design of NOMA combining the characteristics of wireless services can increase network capacity and enable 5G to support high-traffic services.The massive access scenario faces challenge in enabling wireless access of massive energy-constrained devices.Efficient grant-free NOMA resource access scheme is expected to improve connection density.Considering the impact of device energy on connection density,the combination of NOMA and emerging energy-charging technologies can provide energy supply for devices so as to guarantee the connection density improvement brought by NOMA.In this thesis,we study NOMA technology,including the research on user pairing and NOMA-based quality of experience(QoE)improvement of video streaming in enhanced mobile broadband scenario,and the research on through-put improvement of grant-free NOMA system and energy-efficient transmission design in massive access scenario.Based on stochastic geometry,optimization theory and reinforcement learning,we analyze and optimize network performance in terms of coverage probability,spectrum efficiency,access throughput and energy efficiency.Detailed contributions are summarized as follows.(1)To deal with the tradeoff between network capacity improvement and successive interference cancellation(SIC)complexity reduction,we study user pairing schemes in cooperative NOMA networks.With fixed locations of a source and a typical user,the candidate users for pairing are spatially distributed as homogeneous Poisson Point Process.Considering the impact of transmission distance on SIC decoding and network capacity gain,we propose two distance-based user pairing schemes,namely Close-to-Source Pairing(CSP)and Close-to-User Pairing(CUP).To evaluate the performance of proposed schemes in half duplex NOMA and full duplex NOMA systems,expressions for coverage probability and spectrum efficiency are derived using stochastic geometry and Gaussian-Chebyshev quadrature.Numerical and simulation results validate the analysis and reveal that when the distance between the source and typical user is less than the distance threshold,CUP-based full duplex NOMA outperforms non-cooperative NOMA with 11.6%performance gain on sum spectral efficiency.(2)To deal with the QoE improvement of video streaming restricted by network capacity,and the QoE reduction resulting from the mismatching between video quality and data rate,we study QoE-driven resource allocation and video quality adaptation in NOMA networks.A synchronous adaptive bitrate video streaming model is proposed.With the objective of maximizing users'QoE meanwhile reducing the energy consumption,a joint optimization problem of resource allocation and bitrate adaptation is formulated subject to SIC stability constraint and multislot sum rate constraint.With Lagrange relaxation method,the original problem is decomposed into joint subchannel assignment and power allocation(JSAPA)subproblem in each slot,and bitrate adaptation(BA)subproblem for each segment.Further,a joint resource allocation and bitrate adaptation algorithm is proposed based on a modified Gale-Shapley matching algorithm and differ of convex programming.Simulation results reveal that the proposed algorithm improves average bitrate by 15%with insufficient radio resources or with large segment size.(3)To deal with the signaling overhead and computational resource consumption resulting from grant-based access,we propose a deep reinforcement learning(DRL)-based grant-free NOMA resource access framework.To alleviate interference,reduce the computational complexity of DRL and the complexity of SIC,subchannel and device clustering are designed.Then-thcluster of devices compete for the n-th cluster of subchannels with grant-free NOMA.To further reduce intracluster interference,discrete uplink power control is designed to accommodate certain collisions at each discrete received power level.With the objective of maximizing the long-term cluster throughput,a joint optimization problem of subchannel selection and received power level selection is formulated as a partially observable markov decision process,and solved by the proposed DRL-based grant-free NOMA algorithm.Simulation results reveal that when the number of devices is twice and five times that of the subchannels,the proposed algorithm outperforms contention-based grant-free NOMA with 32.9%,156%performance gain on access throughput,respectively.(4)In NOMA MTC networks,insufficient energy supply of MTC devices is a critical issue in limiting access throughput.Benefit by the advantage of energy harvesting(EH)in enabling self-sustainable operation of energy-constrained devices at a low cost,we propose EH-amplify-and-forward(AF)and EH-decode-and-forward(DF)data aggregation schemes.With the objective of minimizing energy consumption,joint optimization problems of power allocation and time division are formulated subject to communication and EH constraints.With optimization theory,the original problems are transformed and finally solved by the proposed EH-AF and EH-DF data aggregation resource allocation algorithms.Simulation results reveal that when the distance between MTC gateway and base station is less than the distance threshold,EH-AF data aggregation scheme saves 42%energy compared with EH-DF data aggregation scheme.With less MTC devices performing NOMA,the distance threshold becomes larger.In conclusion,leveraging NOMA technology,we propose two user pairing schemes for cooperative NOMA,and a resource allocation and video quality adaptation algorithm in enhanced mobile broadband scenario.Besides,we propose a grant-free NOMA resource access framework,and two EH data aggregation schemes and corresponding resource allocation algorithms in massive access scenario.With proposed solutions,the inter-user interference in NOMA system is alleviated,and the goal of exploiting NOMA to improve network capacity and connection density of 5G is achieved.The study in this thesis can provide theoretical insights for the design and application of NOMA in 5G system.