Multi-Agent Reinforcement Learning Based Radio Resource Allocation for Video Transmission in Multi-AP Transmission

The demand for large-capacity, latency sensitive applications such as ultra-high-definition video transmission is increasing in wireless communication systems. In the next-generation wireless local area network (LAN), multi-access point (AP) coordination technique is attracting attention for the purpose of reducing collision probability and improving throughput. In high-demand applications such as ultra-high-definition video transmission, there is a problem that increasing the capacity of wireless communication does not necessarily lead directly to their successful implementation. To accommodate more applications, it is important to enhance video throughput calculated based on whether the requirements for ultra-high-definition video transmission are satisfied. Considering the requirements of video transmission, when multi-APs coordinate and allocate each station (STA) to a radio resource, the number of patterns of when and to which STAs resources should be allocated becomes enormous. As a result, it is difficult to find an optimal resource allocation. Therefore, we propose a radio resource allocation for video transmission, utilizing reinforcement learning (RL). We consider two kinds of RL, which are single agent reinforcement learning (SARL) and multi-agent reinforcement learning (MARL). Our computer simulations demonstrate that MARL can achieve performance comparable to that of SARL while requiring minimal information exchange.