A Learning Approach Towards Power Control in Full-Duplex Underlay Cognitive Radio Networks

Adopting full-duplex (FD) technique in the underlay cognitive radio network (CRN), the secondary user can sense the activity of the primary user (PU) and transmit data simultaneously to enhance spectrum reuse efficiency. However, the sensing accuracy degrades due to the self-interference compared with the discrete sensing and transmission half-duplex network. We employ deep reinforcement learning models to analyze and solve the power control problem with the objective to minimize the adjustment steps in FD underlay CRNs. The proposed power control algorithm can help the secondary transmitter search an optimal transmit power, which satisfies pre-defined quality of service (QoS) requirements while retaining interference to the PU under a threshold. Simulation results show that compared to the benchmark, our method can achieve lower average number of transactions, and reduce more than 53.3% and 98.1% of the computing time and the storage resources, which verifies the effectiveness of our scheme.