A Deep Reinforcement Learning Based Spectrum Access Scheme in Unlicensed Bands
Errong Pei, Yige Huang, Yun Li · 2021
To meet the increasing demand for mobile data traffic, the unlicensed spectrum is used by mobile operators as a complement to the licensed spectrum. Hence, a harmonious and efficient coexistence scheme between LTE and incumbent users such as WiFi in the unlicensed spectrum is urgently needed. At present, advanced artificial intelligence technology is expected to play a key role in future communication systems. Therefore, this paper attempts to introduce the Deep Q-learning framework into the unlicensed spectrum access technology, and further proposes a DRL-based unlicensed spectrum access (DUSA) scheme. In the proposed scheme, based on the proposed access framework, the agent can select the optimal access time and transmission duration by repeatedly interacting with the environment, which can maximize the throughput of the coexistence network while ensuring overall fairness. A large number of simulation results show that the proposed DUSA method can achieve better performance in terms of average rewards, throughput, and fairness than the other baseline methods.