Rank Reversal Avoiding Algorithm for Modern Cooperative Spectrum Sensing

To achieve high reliable of the licensed user detection of cognitive radio, the cooperative spectrum sensing requires a high reliable licensed user detection results from the unlicensed users to perform the cooperative spectrum sensing. Lately, modern cooperative spectrum sensing methods exploit AHP with machine learning models for determining the quality of unlicensed users and the fusion center gathers the local sensing results from only reliable unlicensed users. However, the rank reversal, which is the major issue of AHP that cost a rank reverse, cannot be avoided and addressed. In this paper, we propose a rank reversal avoiding algorithm for the modern cooperative spectrum sensing method by exploiting the epsilon-greedy algorithm for addressing the duplicate information of the local sensing results. From the simulation results, when applying our proposed algorithm to the cooperative spectrum sensing methods, the rank reversal issue can be avoided completely since the value of each duplicated information is expand using epsilon-greedy algorithm under the detected boundary. Moreover, the error of licensed user detection can be reduced once only the local sensing results are only gathered from the reliable unlicensed users. Then, cooperative spectrum sensing performance can be improved by addressing the rank reversal issue.