Resource Allocation Based on Deep Reinforcement Learning for Wideband Cognitive Radio Networks

Abstract

A wideband cognitive radio network relying on energy harvesting is studied under a practically non-linear energy harvesting model. In order to efficiently exploit the harvesting energy and enhance the performance of the secondary users (SUs), an intelligent resource allocation scheme based on deep reinforcement learning is proposed. The scheme intelligently selects the operation mode and the transmit power of SUs as well as allocates the sub-channels to maximize the defined reward function. Simulation results demonstrate the efficiency of our proposed resource allocation scheme.