Reinforcement Learning-Based Joint User Pairing and Power Allocation in MIMO-NOMA Systems.

Area of Science:

• Wireless Communication Systems
• Signal Processing
• Machine Learning Applications

Background:

• Non-orthogonal Multiple Access (NOMA) is crucial for enhancing spectrum efficiency in 5G systems.
• Integrating Multiple-Input Multiple-Output (MIMO) with NOMA further boosts spectral efficiency.
• NOMA systems face challenges with high computational complexity due to dynamic radio channels, hindering efficient resource allocation.

Purpose of the Study:

• To address the computational complexity in MIMO-NOMA systems.
• To develop an efficient scheme for joint user pairing and power allocation.
• To leverage reinforcement learning (RL) for optimizing NOMA performance.

Main Methods:

• Proposed a reinforcement learning (RL)-based joint user pairing and power allocation scheme.
• Utilized Q-learning to simultaneously manage user pairing and power allocation.
• Evaluated the scheme in a multiple-input multiple-output (MIMO)-non-orthogonal multiple access (NOMA) system.

Main Results:

• The RL-based scheme significantly reduces computational complexity compared to traditional methods.
• Achieved a sum rate comparable to exhaustive search (ES) methods.
• Demonstrated effective resource allocation in dynamic radio channel environments.

Conclusions:

• Reinforcement learning offers an effective solution for optimizing MIMO-NOMA systems.
• The proposed Q-learning scheme provides a computationally efficient method for joint user pairing and power allocation.
• This approach maintains high spectral efficiency, making it suitable for future wireless communication systems.