A Low Complexity Solution for Resource Allocation and SDMA Grouping in Massive MIMO Systems

This work investigates the space-division multiple access grouping problem in multiuser massive multiple input multiple output (MIMO). The adopted approach consists in performing firstly the K-means algorithm, that is a classification algorithm well known in machine learning field, to split mobile stations (MSs) into spatially compatible clusters based on the knowledge of the users' spatial covariance matrices. Secondly, it schedules a sub-set of MSs from each cluster thus supporting multiple spatial streams per cluster. Furthermore, the MSs are selected based on a metric that accounts for the trade-off between their spatial channel correlation and channel gain. The corresponding scheduling is optimally solved by using branch and bound (BB) and best fit (BF) algorithms. Moreover, we compare the proposed solutions with the random scheduler that performs clustering and chooses the MSs to compose the groups at random. The simulation results show that the two proposed solutions, BB and BF outperform, the random scheduler. The BB and BF solutions achieve similar capacity performance, but the first has polynomial-time computational complexity while the second a exponential computational complexity.