Edge Computing Resource Allocation Method Based on Federated Reinforcement Learning

With the increase in the number of smart devices, the data generated is growing exponentially, leading to a large amount of data flowing to the cloud and causing congestion in core network communication. Edge computing can effectively reduce the burden of the core network, but there are still challenges in resource allocation between edge nodes and cloud centers. A method based on federated learning and reinforcement learning (FLRL) has been proposed to address this issue, and the server and edge hosts jointly formulate resource allocation strategies, which is beneficial for the global system. Experiments have shown that this method outperforms traditional deep reinforcement learning algorithms in terms of system average utility, average time step, and average resource utilization.