Table 4 Reinforcement machine learning-based user clustering schemes in NOMA

SARSA Q-learning deep reinforcement learning

[48]

Uplink

Base station and users equipped with omnidirectional antennas

Sum rate maximization

IoT

To solve user clustering problem in light and heavy traffic

Modified object migration automation algorithm

[49]

Downlink

Base station knows the CSI of all the users

Sum rate maximization

RIS

To make clusters are in

equal size and perform long term resource allocation

Coalition game approach

[50]

Downlink

Base station knows the CSI of all the users

Spectral efficiency maximization

5G MIMO

Improved cluster beamforming approach due to its fast convergence and flexible cluster size

[51]

Downlink

Base station knows the CSI of all the users

Sum rate maximization

Hybrid MISO

Clustering is performed by

obtaining the optimal solution

[52]

Downlink

Sub carrier assignment must before the

resource allocation of users.

Base station knows the CSI of all the users

Throughput maximization

Device to device

Clustering is based on the basis of coalition and their maximum utility

[53]

Downlink

Number of clusters must be equal to the RF sources in base station channel conditions known by the base station

Sum rate maximization

mm-wave

MISO

Provide low-complexity user clustering mechanism for considering two different cases