DRL-based Dynamic Spectrum Access for Cognitive Radio Networks in Mobility Scenarios

Zhifeng Jiang,Liang Han

DOI: https://doi.org/10.1109/iaecst60924.2023.10503376

2023-01-01

Abstract:Dynamic spectrum access (DSA) is a technology used to address spectrum underutilization. However, most existing DSA studies focus on the impact of spectrum sense results of secondary users (SUs) on access spectrum holes opportunistically but ignore the possibility of spectrum sharing due to users’ mobility. Therefore, this work focuses on the DSA problem for cognitive radio networks in mobility scenarios. Specifically, in each time slot, we design the locations of SUs and primary users (PUs) to be randomly generated within the network coverage, representing the random movement of users. User’s movement causes location changes, and thus we introduce two threshold distances for each SU, enabling SUs to explore opportunities for spectrum sharing with different users at different locations. Then, we use deep reinforcement learning to propose a DSA scheme based on a Double Deep Q-Network (DDQN) for mobility (MD 3 SA). Simulation results show that the performance of the proposed scheme has significantly improved compared to the traditional DDQN-based DSA scheme without mobility consideration (D 3 SA) and random access scheme.