Research On Key Technologies In Non-orthogonal Multiple Access For 5G

In the mobile communication systems,the multiple access techniques are to divide the wireless resources into the orthogonal or non-orthogonal resource blocks,which are then allocated to multiple users according to a certain criteria.Each generation of mobile communication systems have witnessed a “revolution” in terms of their multiple access techniques.Therefore,the multiple access techniques of the physical layer has become an iconic technique of every generation of mobile communication systems.In order to meet the demands of massive connectivity and high throughput for the fifth generation mobile communication(5G)network,non-orthogonal multiple access(NOMA)technique has become a research focus around the world.Sparse code multiple access(SCMA)is a new code-domain NOMA technique,where low-density spreading and multi-dimensional modulation techniques are merged together to directly map incoming data streams to multidimensional complex codewords selected from a predefined cookbook.The optimal codebook set is desgined and selected through the conjugate,permutation and phase rotation.Thanks to the sparsity of SCMA codewords,message passing algorithm(MPA)based receiver is used to recover signals,which can approach the capacity boundary of the multi-user channel.Due to the non-orthogonal sparse coding superposition technique,SCMA technique can support more users than the number of available orthogonal resources.The future massive machine type communications(mMTC)has the characteristics of massive connectivity,small packets,low user data rates and sporadic communication.Grant-free NOMA technique has recently gained significant attention for reducing signaling overhead,latency and terminal power consumption in the mMTC scenario.To accelerate the applications of both SCMA and grant-free NOMA in practice,we investigate several key issues of both techniques.Regarding the low-complexity high-performance detection technique for SCMA system,this thesis starts from two aspects: First,by starting from the improvement of the message delivery method in the iterative process,we can accelerate the convergence rate so that the number of iterations needed to converge is reduced,thereby reducing the computational complexity of the algorithms;Second,reducing the computation amount of the message update operations for per node,that is,to reduce the computational complexity of the algorithms by reducing the computation amount for each iteration.Regarding the signal detection technique for grant-free NOMA system,this thesis proposes a series of new signal detection schemes from the novel perspective of compressive sensing theory,by exploring and exploiting the structured sparsity characteristics of two typical transmission models(frame-wise structured sparsity transmission model and burst-sparsity transmission model)in the mMTC scenarios.To be specific,the contributions of this thesis lie in the following several parts.1)In order to accelerate the convergence rate of the multiuser detection algorithm for SCMA,by starting from the improvement of the message delivery method in the iterative process,we propose two detection algorithms based on the serial message update strategy of resource nodes(RNs)and user nodes(UNs).In this way,the updated messages can join the message propagation immediately in current iteration so that the convergence rate is accelerated.Furthermore,from the perspective of high-speed detection implementation,the serial scheduling baesed detection algorithm is extended to the parallel form with the aid of a group method,which not only inherits the advantage of the fast convergence,but also greatly improves the detection speed.In order to construct a more reasonable message update order to further accelerate the convergence rate,we first propose a serial scheduling MPA based on weight of user node,where the user node scheduling order is first selected based on the maximum number of newly available message updates in the factor graph.Then,a detection algorithm based on residual-aided dynamic message scheduling strategy is proposed,which exploits the dynamic selection of the UN-to-RN message which has the largest residual,whenever messages are updated sequentially.Finally,we verify the proposed serial scheduling based detection algorithms through the simulation results.2)The main computational complexity of MPA is due to the message update operations at resource nodes,which grows exponentially with the effective degree of the resource node.To reduce the computational complexity,we propose a Gaussian approximation aided MPA.The basic idea is to apply Gaussian approximation to part of the connections so that the factor graph can be simplified by selecting some neighbors for each edge,which results in a dynamic graph.Then,we propose a compressive sensing aided MPA detector,combined MPA and compressive sensing technique.Specifically,this detector employs a sparse error recovery algorithm to refine the estimate of a symbol vector obtained by the MPA with a few iterations.In order to strike a balanced trade-off between detection performance and computational complexity,a joint SCMA and MIMO detection scheme is proposed for downlink MIMO-SCMA systems,by making full use of the graphical representations of MIMO channels and SCMA codewords.More detailedly,we construct a joint sparse factor graph combining the single graph of MIMO and SCMA,and then design the corresponding virtual SCMA codebooks.Based on them,MPA can be directly applied to reconstitute the transmitted data bits.3)For the frame-wise structured sparsity transmission model in the grant-free NOMA system,by exploring the block sparisty inherent in the structured sparsity structure,we propose a signal detection scheme that directly reconstructe the signals in a whole frame and does not require the prior knowledge of user sparsity level under the framework of block compressive sensing.Firstly,for the joint user activity and data detection,we propose a threshold aided signal detection algorithm that can approach the near-optimal performance and a cross validation aided signal detection algorithm that does not require any prior knowledge,respectively.Then,for the situation that the channels for all users are not known a priori,a joint channel estimation and data detection scheme is proposed,by considering the inherent frame-wise structured sparsity of the pilot and data phases in the entire frame.Additionally,we provide the convergence analysis and the the complexity analysis of the two proposed algorithms under the framework of compressed sensing theory.4)For the burst-sparsity transmission model in the grant-free NOMA system,introducing a parameter indicating the quality of the prior support,we proposed a prior-information aided signal detection algorithm.Specifically,this algorithm exploits the prior support adaptively based on the quality information,so as to fully explore and exploit the intrinsically temporal correlation of active user support in several continuous time slots.To tolerate possible model mismatch,we further propose a more robust signal detection algorithm to combat incorrect prior support quality information,which adaptively exploits the prior support based on the corresponding support quality information in a conservative way.In addition,the convergence analysis as well as the complexity analysis of the two proposed algorithms are also provided under the framework of compressed sensing theory.As expected,the above mentioned theoretical results and algorithms can develop and improve the theoretical framework,technical solutions and algorithms of NOMA,and promote the application of key technologies of NOMA in future mobile communication systems.