Joint Deep Reinforcement Learning and Unsupervised Learning for Channel Selection and Power Control in D2D Networks

Area of Science:

• Wireless Communication
• Network Resource Management
• Artificial Intelligence in Telecommunications

Background:

• Device-to-device (D2D) communication offers a solution to spectrum scarcity in 5G networks.
• Shared channels in D2D networks lead to significant interference, hindering capacity and spectral efficiency.
• Existing resource allocation methods often rely on centralized control, requiring extensive network information.

Purpose of the Study:

• To develop a distributed resource allocation algorithm for D2D networks.
• To mitigate interference and enhance wireless spectrum utilization.
• To improve network capacity and spectral efficiency in 5G systems.

Main Methods:

• A deep Q-network (DQN) was employed for distributed channel allocation in dynamic environments.
• An unsupervised learning-based deep neural network was utilized for optimized power control.
• The proposed algorithm operates distributively, using local information for channel selection and power control.

Main Results:

• The proposed algorithm effectively reduces interference in D2D user pairs.
• It significantly enhances network capacity and wireless spectrum utilization.
• Simulation results demonstrate superior performance compared to traditional centralized and distributed algorithms in terms of convergence speed and transmit sum-rate maximization.

Conclusions:

• The integration of DQN and unsupervised learning provides an effective distributed solution for D2D resource allocation.
• The algorithm achieves higher spectral efficiency and faster convergence than conventional methods.
• This approach offers a scalable and efficient strategy for future wireless communication systems.