Spectrum Sensing-Enhanced Federated Multi-Armed Bandit for Cognitive Radio

This repository contains the code and findings from the research paper titled "Spectrum Sensing-Enhanced Federated Multi-Armed Bandit for Cognitive Radio" by Kanghee Lee and Jungmin So, conducted under the Department of Computer Science and Engineering at Sogang University.

Abstract

This study enhances the Federated Multi-Armed Bandit (FMAB) algorithm for cognitive radio (CR) systems by introducing uncertainty in frequency band rewards and using energy detection for spectrum sensing. The enhanced FMAB algorithm demonstrates improved spectrum access, reduced exploration costs, and faster convergence in uncertain CR scenarios.

Contents

1. Code Implementation
2. Simulation Results
3. Comparative Analysis between Fed1 UCB and Fed2 UCB

Key Findings

• Fed2 UCB (client cooperation) outperforms Fed1 UCB (independent client behavior) in reducing regret and communication costs.
• Randomized mean rewards and spectrum sensing lead to more efficient and effective decision-making in dynamic spectrum conditions.