| With the explosive growth of information,the shortage of spectrum resource is one of the main problems hindering the development of communication nowadays.Cell-free massive multiple input multiple output(MIMO)combines the advantages of traditional cellular MIMO and distributed MIMO,and has higher coverage and stronger capability of interference suppression,so it has become one of the potential technologies for B5 G and 6G.The degrees of cooperation among the access points(APs),pilot assignment and power control in cell-free massive MIMO system are researched in this thesis,details are as follows:The implementations of cell-free massive MIMO system are divided into three levels according to the degrees of cooperation among the APs,which are fully local processing,partial local processing and fully centralized processing.In full local processing,channel estimation and data preprocessing are performed locally by each AP,and then the estimates of data signals obtained by data preprocessing are sent to the central processing unit(CPU),the CPU performs the final decoding by simply taking the average of the estimates.In partial local processing,the estimates of data signals sent by each AP are combined linearly at the CPU according to the optimum weight vector.In fully centralized processing method,the pilot signals and data signals received by APs are sent directly to the CPU,and the channel estimation and data processing are all performed at the CPU,so the APs can be regarded as repeaters.The uplink spectrum efficiency(SE)under these three processing methods are analyzed in this thesis,and performance of the system when using minimum mean square error(MMSE)and maximum ratio combining(MRC)processing under different degrees of cooperation among the APs are compared.Finally,under the same data processing method,fully centralized processing is better than partial local processing,partial local processing is better than fully local processing,and the least fronthaul link resources are needed in fully centralized processing method.To reduce pilot contamination,a pilot assignment method based on an improved K-means algorithm is proposed in this thesis.Based on the traditional K-means algorithm,the steps of substantial centroid selection,solitary centroid transfer and cluster density control are added,so that the output of the algorithm is more consistent with the requirements of the system model in this thesis.The result of the simulation shows that compared with the traditional random pilot assignment and random grouping pilot assignment,the sum SE of the system can be improved effectively by the method proposed in this thesis.In order to improve the performance of the system furtherly,a pilot assignment method based on differential evolution algorithm is proposed in this thesis.Through the processes of crossover and mutation,the pilot assignment is evolved towards the optimal assignment.Simulation result shows that the sum SE of the system can be maximized effectively by this method.For the power control,an improved firefly algorithm is adopted to optimize the transmission power of uplink data of each user in order to maximize the sum SE of the system,and it is compared with full power transmission and particle swarm optimization(PSO)algorithm.Simulation result shows that the power control algorithm proposed in this thesis can not only improve the sum SE of the system effectively,but also have faster rate of convergence.Finally,the pilot assignment algorithm and the power control algorithm proposed in this thesis are combined to optimize the system jointly,which greatly improves the performance of the system. |
