Massive machine-type communications(mMTC)is one of the most important solutions to support massive connections in the 5th Generation(5G)mobile communication system.It’s characterized by are large-scale device connections,low-latency transmission and sporadic communications.Driven by the Internet-of-Things(Io T),a grant-free Non-Orthogonal Multiple Access(NOMA)is proposed in 5G system,which can significantly reduce the signaling overhead,transmission delay and user terminal power consumption.Due to the lack of dynamic resource allocation process in the uplink grant-free NOMA system,the detection of active users and their transmitted data at the base station has become an important research direction in the mMTC system.The sporadic communication of the mMTC system makes active users show sparseness.At present,compressed sensing(CS)theory is widely used to realize active user detection.In this thesis,aims to explore a detection algorithm with better reconstruction performance and in line with the characteristics of the actual system based on the existing CS-based multi-user detection algorithm.The main research work of this thesis is as follows:1.Aiming at the dynamic multi-slot sparse detection model of the mMTC system,a threshold-assisted adaptive dynamic compressed sensing(TA-ADCS)multi-user detection algorithm is designed in this thesis.The classic dynamic compressed sensing(DCS)algorithm improves the detection performance by using the time correlation between adjacent time slots.However,the DCS algorithm needs to know the user sparsity in advance,which does not conform to the characteristics of the actual communication system.In the lack of user sparsity,the proposed algorithm first introduces an adaptive threshold assisted strategy to select active users to improve the accuracy of atom selection.Secondly,the power function variable step method is used in the sparsity estimation process,according to the iteration of residual signal,the step size of sparsity estimation is dynamically adjusted to improve the accuracy and speed of signal estimation.Finally,the termination conditions of the iteration algorithm are optimizedto ensure the performance of sparse signal reconstruction.The simulation results show that the proposed algorithm can adaptively obtain user sparsity,and realize the joint detection of active users and their data with lower computational complexity,and has better detection performance than the classic CS algorithm.2.Taking into account the actual situation that users can randomly access or leave the mMTC system,a hybrid multi-slot sparse detection model is adopted in this thesis,focusing on the analysis of the characteristics of the active user support set under the hybrid sparse model,it can be divided into a common support set and a dynamic support set,We design a gradient information-based step-by-step sparse adaptive matching pursuit(GI-SSAMP)algorithm.The proposed algorithm consists of two parts,namely,the common support set estimation and the dynamic support set estimation.In order to improve the accuracy of the common support set estimation,the hybrid sparse model is converted to a block structure sparse model.Then,the estimation result of common support set is taken as the prior information of the dynamic support set estimation.In the hybrid sparse model,the detection is performed time slot by time slot,and the active users and their transmitted data in each time slot are finally obtained.Considering the high complexity of the greedy algorithm using the least square to estimate the signal,the GI-SSAMP algorithm uses the gradient information iteration based on the error function to replace the matrix inversion,which effectively reduces the computational complexity of the algorithm.The simulation shows that,compared with most CS-based multi-user detection algorithms,the algorithm combining the stepwise detection strategy with the gradient iteration greatly improves the detection performance. |