An Effective Spectrum Handoff Based on Reinforcement Learning for Target Channel Selection in the Industrial Internet...

Area of Science:

• Wireless Communications
• Cognitive Radio Networks
• Resource Management

Background:

• Wireless spectrum overcrowding necessitates efficient resource allocation.
• Dynamic Spectrum Access (DSA) and Cognitive Radio (CR) offer solutions for better channel utilization.
• Spectrum Handoff (SH) is crucial for CR but can degrade performance if not optimized.

Purpose of the Study:

• To develop an optimized Spectrum Handoff (SH) policy for Cognitive Radio (CR) networks.
• To enhance channel selection strategy (CSS) for improved network performance and reduced latency.
• To integrate effective spectrum sensing and channel selection for fair resource allocation.

Main Methods:

• Implemented two reinforcement learning (RL) algorithms within the channel selection strategy (CSS).
• Utilized one RL algorithm for predicting future channel occupancy and another for assessing channel quality.
• Developed a method to mask linearly dependent and irrelevant state elements to accelerate learning convergence.

Main Results:

• Achieved significant reductions in communication latency.
• Demonstrated remarkable improvements in network throughput performance.
• Outperformed conventional approaches in dynamic spectrum access scenarios.

Conclusions:

• The proposed SH policy effectively integrates RL-based CSS and spectrum sensing.
• Optimized channel selection using RL leads to enhanced network efficiency and reduced delays.
• This approach provides a robust solution for fair and reliable resource allocation in crowded wireless environments.