Compressive Sensing Multi-user Detection For Uplink Grant-free NOMA

With the commercialization of 5th Generation mobile networks(5G),massive Machine Type Communications(m MTC)will become the dominant wireless communication in the future.Grant-Free non-orthogonal multiple access technology is an organic combination of Non-Orthogonal Muliple Access(NOMA)technology and GrantFree(GF)transmission technology.It can meet m MTC's requirements for connection volume,signaling overhead,and transmission delay.However,the base station cannot know the user's activity status,which brings certain challenges to uplink multi-user detection.Fortunately,because the user activity in the m MTC scenario is sparse,the problem of Grant-Free NOMA uplink multi-user detection can be transformed into the reconstruction of sparse signals,then the knowledge of compressed sensing is used to solve it.Therefore,this thesis studies the multi-user detection algorithm based on compressive sensing for Grant-free NOMA uplink.The main contents are summarized as follows:1.Sparse adaptive matching pursuit multi-user detection algorithm assisted by noise energyIn the static Channel State Information(CSI)environment,for the number of active users in the uplink of Grant-Free NOMA system is unknown,this thesis proposes a sparse adaptive matching pursuit multi-user detection algorithm assisted by noise energy(NEASAMP).Based on the traditional SAMP algorithm,the NEA-SAMP first uses the generalized Dice coefficient matching criterion to calculate the correlation coefficient,thereby improving the accuracy of the active user support set.Then it sets the iteration termination threshold based on the noise energy to avoid overestimation and underestimation of the number of active users.Finally,a variable step size mechanism of fast approach with large steps,accurate approach with small steps is introduced to ensure accuracy and increase speed.Simulation results show that NEA-SAMP has better detection success rate,bit error rate,and iterative times than the traditional OMP and SAMP algorithms.2.structure sparse adaptive matching pursuit multi-user detection algorithm assisted by cross-validation and error estimationIn the dynamic CSI environment,this thesis uses the time correlation of user activity status to convert the multi-slot structure sparse model to a block structure sparse multiuser detection model,and converts the multi-slot detection to a single-slot detection.Based on this,considering the situation that the number of active users in the uplink base station of Grant-Free NOMA system is unknown,a structural sparse adaptive matching pursuit algorithm assisted by cross-validation and error estimation(CV-EE-SAMP)is proposed.Specifically,the algorithm uses cross-validation to improve the traditional SAMP algorithm,and makes preliminary estimates of the number of active users and user signals.Then use the homotopy method to estimate the detection error.Finally,the error estimation value is used to correct the initial user signal estimated value to improve the detection accuracy.Simulation results show that CV-EE-SSAMP has improved bit error rate performance compared with SAMP and CVA-BSASP.Although the error rate performance of NEA-SAMP and CV-EE-SSAMP is basically the same,CV-EE-SSAMP does not require prior information such as known noise energy or signal-to-noise ratio.Therefore,the practicability of CV-EE-SSAMP is better than NEA-SAMP.