Deep Reinforcement Learning for Dynamic Spectrum Access in the Multi-Channel Wireless Local Area Networks
Sovit Bhandari, Navin Ranjan, Yeong-Chan Kim, Hoon Kim · 2022
In recent years, the rapid proliferation of wireless local area networks (WLANs) has led to a scarcity of radio spectrum. Dynamic Spectrum Access (DSA) is considered a promising technology to address the increasing shortage of radio spectrum and improve its utilization. DSA technique effectively utilizes the radio spectrum by switching between different networks. However, most conventional DSA techniques do not consider the correlation between multiple channels and require network information in advance to make decisions. Due to recent advances in reinforcement learning, a deep Q-network (DQN) based method is proposed in this paper to solve the problem of correlated multi-channel DSA with unknown system dynamics. The performance of the DQN-based method is quantified based on the successful packet transmission, packet collisions, and channel utilization.