Uncertainty-Calibrated Deep Cooperative Spectrum Sensing and Channel Access Optimization for Cognitive Radio Networks

Abstract

This has created serious spectrum congestion in most frequency bands due to the growing demand of wireless communication services. The conventional policies of the static spectrum allocation result in inefficient spectrum usage since some geographical areas and time periods have unused frequency bands in a certain region. The solution to this issue is the cognitive radio technology which allows secondary users to use the spectrum dynamically without disrupting primary users who have been licensed. Accurate spectrum sensing is one of the basic conditions that such dynamic access requires since it enables cognitive radio devices to discover whether a channel is occupied. Traditional spectrum sensing schemes like energy detection and matched filtering fail to work well in the presence of low signal to noise ratio or in fading and shadowing conditions. Cooperative spectrum sensing makes more accurate the sensing results because more than one radio may share the sensing information, however, it presents new difficulties associated with decision fusion and imprecision in sensing outcomes. This paper suggests an uncertainty-calibrated deep cooperative sensing system that combines the deep neural network-based signal classification with uncertainty prediction to make reliable channel access decisions. Simulated radio signal conditions are created to test the performance of the proposed methodology without having to use real-world data. Simulation findings have shown that the developed method greatly enhances the detection accuracy and also minimises the detected false alarms relative to traditional methods of sensing. The findings show that uncertainty estimation combined with cooperative sensing can enhance the reliability of cognitive radio networks that work in dynamically changing wireless environments.