Intimacy-based Resource Allocation for Network Slicing in 5G via Deep Reinforcement Learning

In view of the development of emerging IoT applications driven by artificial intelligence, the fog radio access network has recently been introduced to fifth generation (5G) wireless communications to utilize available fog resources. Due to the laten-cy limitation of the cloud, we consider prioritizing the allocation of limited resources of fog nodes on the edge to requests. In this article, we describe the process of the edge controller service request as a Markov decision process. In order to maximize resource utilization while guaranteeing the qual-ity of service, we devise a novel intimacy-based deep reinforcement learning algorithm for network slicing in 5G networks. It establishes effective con-nections between nodes by judging the intimacy between nodes based on the resource status of neighbor nodes and learned experience. The sim-ulation results show that our method outperforms the state of the art in terms of utilization and delay.