Study on the Spectrum Sharing Algorithm Using Reinforcement Learning

In dense network environments, channel resources consumed by traditional media access control (MAC) methods increase. In this paper to reduce consumed channel resources, frequency sharing algorithm using reinforcement learning is developed that optimize channel usage according to current wireless communication environment. The use of proposed frequency sharing algorithm can address the reduction in channel usage due to the increasing overhead of the Backoff algorithm in a dense environment while down-link (DL, Down Link) data transfer at 802.11ax and It can also improve the threshold that media can be accessed through only one channel in CB(Channel Bonding). This paper deals with the implementation of 802.11ax down link environment wireless communication system and DQN(Deep Q-Network) structure and parameters, which will be used as frequency sharing algorithm within the system. Performance of trained reinforcement learning model in wireless communication system is verified and by changing the parameters, we compared the training results and identified the meaning of the changed parameters in frequency sharing algorithm.