Research On Pilot Allocation Scheme Based On Machine Learning In Massive MIMO System

Massive multiple-input multiple-output(MIMO)technology,as one of the key technologies of 5th Generation Mobile Networks(5G),deploys hundreds to thousands of antenna arrays at base station.In order to obtain a higher degree of spatial freedom,thereby effectively improving spectrum efficiency and energy efficiency.In the Time Division Duplex(TDD)mode,channel reciprocity is used to perform channel estimation on the uplink channel to obtain CSI,which often depends on the orthogonality of the uplink pilot signal.However,the channel coherence time is limited in an actual communication system,and the number of orthogonal pilots in the system is limited.Therefore,users in different cells inevitably reuse the same pilot,thereby causing pilot pollution.At present,pilot pollution has become a bottleneck problem that restricts the performance of massive MIMO system.Pilot allocation is one of the effective methods to reduce pilot pollution in massive MIMO system.Therefore,this paper will combine the algorithms in machine learning and conduct research from the perspective of pilot allocation.The main work is as follows:(1)Taking a single target cell as the research object,the pilot allocation of users in the target cell,which is considered as a one-to-one matching problem between users and pilots.Aiming at improving the uplink reach and speed of strong users,while ensuring the quality of service(QoS)requirements of weak users in the system,and improving the overall system and speed as the goal,a low-complexity-based matching algorithm was proposed.Packet pilot allocation scheme.The large-scale fading coefficient is used to measure the channel quality of users,and the target cell users are divided into weak user groups and strong user groups according to the user channel quality,and pilot allocation is performed on the weak user groups using the minimum-maximum matching method.With the introduction of the Hungarian algorithm,a pilot allocation method that can guarantee the fairness of the strong user group is designed.Simulation results show that the system performance of the proposed scheme achieves lower complexity and higher performance,compared with pilot allocation schemes for user grouping,the smart pilot allocation scheme,and the pilot allocation scheme based on Hungarian algorithm.(2)Taking multi-cell as a research object,the pilot allocation of multi-cell users is regarded as a problem of coloring the vertices of the graph one by one.Combining the advantages of graph coloring theory and water injection algorithm,a pilot allocation optimization scheme,which based on weighted graph coloring is designed to improve the overall system rate.First,the user's maximum uplink signal-to-interference and noise ratio is taken as the objective function.Then,the user's degree of pollution is reflected by the edge weight.The larger the edge weight is,the more serious the pilot contamination of the two users is.The users are regarded as vertices and the pilots are regarded as color.The degree of mutual interference between users in different cells is regarded as edge weight,and an edge weight interference graph is constructed.Finally,the water injection algorithm is used to preferentially allocate the pilot with the least pollution to the user corresponding to the smallest edge weight.Simulation results show that the proposed scheme achieves higher achievable sum rate in the system than the weighted-graph coloring pilot allocation scheme and the grouping pilot allocation scheme based on matching algorithm.