کنفرانس مهندسی برق ایران

صفحه اصلی / سی و دومین کنفرانس بین المللی مهندسی برق

Multi-agent H-Learning Based Cooperative Spectrum Sensing for Cognitive Radio Networks

نویسندگان :

Elaheh Karimpour Fard¹ Mahdi Nouri² Hamid Behroozi³ Sima Sobhi-Givi⁴

1- دانشگاه صنعتی شریف 2- دانشگاه صنعتی شریف 3- دانشگاه صنعتی شریف 4- دانشگاه ارومیه

کلمات کلیدی :

Cognitive radio network،Cooperative spectrum sensing،Massive MIMO،Multi-agent reinforcement learning،Model-based H-learning،Gaussian kernel reward function

چکیده :

Fast and accurate methods for spectrum sensing (SS) are the key elements in cognitive radio networks (CRNs) that achieve high SS. This paper proposes a reinforcement learning (RL) scheme for secondary users (SUs) in a massive multipleinput multiple-output (MIMO) CRN to determine occupation and emptiness of scanned wideband channels. In the proposed H-learning method, the clustered neighboring SUs collect data to check the status of channels that improve reliability of measured sensing. Then, they share information about different frequency channels with other clusters and the status of channel occupation is determined for all SUs. This model-based learning method increases finding free channels and enhances spectral efficiency (SE) in presence of large probability of missing detection. Furthermore, each SU learns the presence pattern of primary user (PU) and gather a dynamic scan priority list to decrease scan overhead and access delays. Simulation results demonstrate the multi-agent reinforcement modified H-learning (MH-learning) method can achieve better performance in terms of PU usage percent, number of attempts, call block rate, and probability of detection in comparison to H-learning and some state-of-art learning methods.