ScholarWorks@중앙대학교

초록

—In spectrum sensing, secondary users (SUs) often face limitations in their sensing capabilities, making cooperation among SUs crucial. By leveraging collaboration, we aim to maximize comprehensive awareness of primary channel states while minimizing sensing overhead. To achieve this, we develop an adaptive algorithm that utilizes SU mobility and directional antennas, enabling operation in dynamic environments where prior network knowledge is unavailable. This study proposes a two-stage learning approach. In the first stage, a PU location estimation (PLE) algorithm assists SUs in identifying relevant sensing targets. In the second stage, a multiagent reinforcement learning-based algorithm is introduced to mitigate sensing collisions, thereby enhancing channel state awareness. Simulation results demonstrate that the proposed scheme outperforms baseline algorithms in convergence speed, sensing collision rate, and channel state awareness, offering an efficient solution for cognitive radio networks. Specifically, the proposed approach achieves up to an 81% and 24% increase in comprehensive channel awareness, and a 53% and 13% reduction in sensing overhead compared to the random scheme without and with PLE, respectively.

키워드