Dynamic Spectrum Sharing and Aggregation Scheme Based on Deep Reinforcement Learning
Tianqi Sheng, Wensheng Zhang, Wenjiao Ding, Jian Feng Sun, Cheng‐Xiang Wang · 2022 International Wireless Communications and Mobile Computing (IWCMC) · 2022
A novel spectrum sharing and aggregation scheme based on the maximum entropy actor-critic (MEAC) algorithm is proposed to solve the spectrum shortage problem in dynamic spectrum access (DSA). The spectrum sharing and aggregation problem is modeled as a three-state Markov model, where secondary users (SUs) try to access multiple idle channels. In a time slot, the SUs with the spectrum sensing, sharing, and aggregation capabilities select available spectrum slots and channels through spectrum sharing and aggregation. The actor-critic algorithm is used to construct the spectrum framework, and a novel maximum entropy (ME) scheme is proposed to achieve an optimal spectrum sharing and aggregation policy. The ME scheme can include more exploration and ensure that the output action is more stochastic than other schemes. Simulation results indicate that the proposed scheme can achieve better performance than Deep Q-Network (DQN) algorithm.