Spectrum-efficient user grouping and resource allocation based on deep reinforcement learning for mmWave massive...

Area of Science:

• Wireless communication networks
• Signal processing
• Machine learning applications

Background:

• Millimeter-wave (mmWave) massive multiple-input multiple-output non-orthogonal multiple access (MIMO-NOMA) is crucial for 6G.
• Increasing users and antennas in mmWave massive MIMO-NOMA systems pose challenges in interference suppression and resource allocation.

Purpose of the Study:

• To propose a spectrum-efficient and fast-converging deep reinforcement learning (DRL) framework.
• To optimize user grouping, subchannel allocation, and power allocation in mmWave massive MIMO-NOMA systems.

Main Methods:

• An enhanced K-means algorithm for user grouping to reduce interference and accelerate convergence.
• A dueling deep Q-network (DQN) for subchannel allocation, improving convergence speed.
• A deep deterministic policy gradient (DDPG)-based algorithm for power resource allocation to enhance system sum-rate.

Main Results:

• The proposed DRL framework demonstrates superior convergence performance compared to other neural network-based algorithms.
• The scheme achieves higher system capacity than greedy, random, RNN, and DoubleDQN algorithms.
• The DDPG-based power allocation avoids performance loss from power quantization.

Conclusions:

• The developed DRL framework effectively addresses resource allocation challenges in mmWave massive MIMO-NOMA systems.
• The proposed methods significantly improve convergence speed and system capacity.
• This approach offers a promising solution for efficient 6G wireless communication.