A Graph Convolutional Network-Based Deep Reinforcement Learning Approach for Resource Allocation in a Cognitive Radio...

Area of Science:

• Wireless communication networks
• Spectrum management
• Machine learning applications in telecommunications

Background:

• Explosive traffic growth exacerbates spectrum scarcity, necessitating advanced techniques like cognitive radio (CR).
• Effective CR resource allocation is key for spectrum sharing and mitigating co-channel interference (CCI).

Purpose of the Study:

• To propose a joint channel selection and power adaptation scheme for underlay cognitive radio networks (CRNs).
• To maximize secondary user (SU) data rates while ensuring primary user (PU) quality of service (QoS).

Main Methods:

• Modeling CRNs as dynamic graphs with random walks to simulate user movement.
• Estimating channel state information (CSI) using user distance distribution from graph topology.
• Employing graph convolutional networks (GCN) for interference feature extraction.
• Designing an end-to-end learning model integrated with deep reinforcement learning (DRL) for resource allocation.

Main Results:

• The proposed scheme effectively manages radio resources in dynamic CRN environments.
• Simulation results demonstrate the feasibility and convergence of the deep reinforcement learning approach.
• Significant performance improvements in data rate maximization and QoS guarantee were observed.

Conclusions:

• The joint channel selection and power adaptation scheme offers a robust solution for underlay CRNs.
• The integration of graph theory and deep learning provides an effective framework for complex radio resource management.
• This approach significantly enhances spectrum utilization efficiency and network performance.