A User Clustering Method for Large-Scale MIMO System

Abstract

A user clustering method under a massive MIMO system, comprising the steps of: S1. searching for noisy users, and eliminating them; S2. obtaining the three-dimensional features of the users, and calculating a similarity matrix of the three-dimensional features; S3. according to the similarity matrix, Use the AP algorithm to cluster users; S4. Use Q-learning to learn and predict the status of user clusters to obtain the final clustering results. The invention can perform effective clustering processing on users, thereby reducing the dimension of the system, reducing the processing dimension to a certain extent, reducing interference, and improving resource utilization.

Description

technical field [0001] The invention belongs to the field of wireless communication, relates to a 5G (5th-generation) mobile communication system, and specifically relates to a user clustering method under a massive MIMO system. Background technique [0002] While the rapid development of wireless communication technology promotes the continuous improvement of network infrastructure, it also leads to explosive growth in the number of mobile users and the scale of related industries. Existing spectrum resources are limited, and in many applications, it has been unable to meet the increasing demand for capacity. In order to solve the problems raised above, Massive MIMO, as one of the most critical technologies of the fifth generation (5G) mobile communication, was proposed in "5G Wireless Technology Architecture" and "5G Concept White Paper". In a multi-user Massive MIMO system, spatial correlation has a significant impact on the improvement of system performance, that is, th...

AI Technical Summary

Problems solved by technology

[0003] Among them, the k-means clustering algorithm [1] does not have stable selection of initial points and is randomly selected, which causes instability in clustering results; secondly, although hierarchical clustering [5] does not need to determine the number of classifications, once Once a split or merge is executed, it cannot be corrected, and the quality of clustering is limited; moreover, the DBSCAN algorithm [6] is a density-based user grouping algorithm, and it is necessary to select an appropriate radius and a threshold for the minimum number of users. If it is not selected properly, it will Affect clustering results

Finally, the FCM algorithm [7] is sensitive to the initial cluster center and needs to manually determine the number of clusters, which is easy to fall into a local optimal solution;

Images