Hybrid CNN-LSTM-Based Cognitive RF-AI Architecture for Intelligent Spectrum Sensing in 6G Networks

Abstract

The rapid growth of ultra-dense wireless devices and heterogeneous communication services in sixth-generation (6G) networks creates significant challenges for efficient spectrum utilization and interference management. Conventional spectrum sensing techniques suffer from limited adaptability and reduced detection accuracy under highly dynamic and congested radio environments. To address these issues, this paper presents a Cognitive RF-AI Co-Design framework for intelligent spectrum sensing in dense 6G wireless environments. The proposed architecture integrates reconfigurable radio-frequency (RF) front-end modules with artificial intelligence (AI)-based learning models to enable adaptive and autonomous spectrum awareness. A hybrid deep learning framework combining Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks is employed to extract spatial and temporal characteristics from RF signals for accurate spectrum occupancy prediction. Reinforcement learning mechanisms further optimize sensing parameters and channel selection strategies in real time. The RF-AI co-design approach enhances signal detection capability under low signal-to-noise ratio conditions while reducing false alarm probability and sensing latency. Simulation results demonstrate that the proposed system achieves 98.4% spectrum sensing accuracy, 96.8% channel occupancy prediction accuracy, a 22.5% reduction in sensing delay, and approximately 18.7% improvement in spectrum utilization efficiency compared with conventional cognitive radio approaches. The proposed framework provides an intelligent and scalable solution for dynamic spectrum management, enabling reliable and energy-efficient communication for future dense 6G wireless networks.