With the recent development of communication technology, networked control systems in which the control target and the controller are remotely networked have been attracting attention. This networked control system has the advantage of simplifying installation and maintenance because the components of the system can be distributed wirelessly. On the other hand, the control performance is degraded due to data delay, missing data, and limited communication capacity caused by the transmission. Therefore, it is necessary to develop a communication method that can take control performance into account.

 In this study, we pay particular attention to the communication capacity limitation and propose a communication reduction method that takes control performance into account. As a communication reduction method, we adopt a "scheduling" method, in which a transmission element is selected at each control cycle. The feature of this study is that this scheduling rule is derived by reinforcement learning. The method using reinforcement learning does not require a model of the control target to derive the scheduling rule. Therefore, this method can be applied regardless of the type of the control target. The effectiveness of the scheduling method using reinforcement learning will be clarified through simulations and experiments.