Research On Sparse Multi-user Detection Algorithms In Massive Machine Type Communication Systems

With the rapid development of the Internet of Things technology,the industry has gradually begun to study,design and build next-generation wireless communication system to meet the growing needs of life.Massive Machine Type Communications(mMTC)is one of the three major application scenarios in the next generation of wireless communication system.The main features of the mMTC system include massive device access,sporadic communication,lower rates,and short data packets.Due to these characteristics of the mMTC system,the multi-user detection of the mMTC system is similar to the sparse signal reconstruction in the theory of compressed sensing(CS).Therefore,the CS-based signal reconstruction methods can be applied to multi-user detection in the mMTC system.In addition,sparsity-aware successive interference cancellation(SA-SIC)and its improved algorithms are also feasible methods to detect these low-data-rate multi-user signals.This thesis considers the single-slot model of the uplink grant-free mMTC system,where multi-user detection algorithms are investigated.The main contents are as follow:Aiming at the uplink grant-free mMTC system,the CS-based signal detection algorithms are mainly addressed.In recent years,the industry has applied the CS-based signal reconstruction methods to the mMTC system to achieve multi-user detection.The generalized orthogonal matching pursuit(gOMP),as a typical CS algorithm,gradually approximates the original signal in a greedy iterative manner.This algorithm is well implemented for sparse signal reconstruction with the advantages of simplicity and easy implementation.However,it is found that the algorithm has the disadvantage of undetermined indices number,which leads to the reconstruction performance of the algorithm being affected.This thesis first determines the optimal number of indices theoretically.Then,the multiple optimal indices aided generalized orthogonal matching pursuit(MOIA-gOMP)algorithm is obtained.At last,the performance of MOIA-gOMP algorithm is validated in simulation.The simulation results show that compared with other greedy pursuit algorithms and gradient projection for sparse reconstruction algorithms,the MOIA-gOMP algorithm has better performance.And for the mMTC system configured with different parameters such as the number of active users and the number of subcarriers,the MOIA-gOMP algorithm shows better multi-user detection performance than other algorithms mentioned in this thesisTowards the uplink grant-free mMTC system,the signal detection algorithms based on SA-SIC are also coped with.The CS technology lacks methods for processing sparse signals under the constraints of finite modulation set.The industry has proposed various typical methods to convert the compressive sensing multi-user detection(CS-MUD)into the SA-SIC detection problems.The traditional SA-SIC method lacks the sorting mechanism,resulting in the error propagation phenomenon.The error accumulation may deteriorate the performance of SA-SIC detection.Activity-aware sorted orthogonal triangular decomposition(A-SQRD)is an improved algorithm of SA-SIC.The detection order is jointly determined by the channel gain and the devices' active probability in the A-SQRD algorithm,thereby to mitigate error propagation and improve performance.The A-SQRD algorithm can achieve good performance compared to the SA-SIC algorithm.The performance of the A-SQRD algorithm can still be improved through further interference cancellation.The improved activity-aware sorted orthogonal triangular decomposition(IA-SQRD)algorithm is proposed in this thesis.The IA-SQRD algorithm utilizes the final solution of the traditional A-SQRD algorithm as the initial solution and adds an iterative interference cancellation operation to further improve detection performance.The simulation results show that compared with the A-SQRD algorithm,the IA-SQRD algorithm proposed in this thesis can achieve 3 dB performance gain without increasing the computational complexity significantly when the bit error rate(BER)is 2.5×10-2.In addition,for the mMTC system configured with different parameters such as active probability or spreading sequence length,the IA-SQRD algorithm shows better multi-user detection performance than other algorithms mentioned in this thesis.