Slicing Framework for Service Level Agreement Guarantee in Heterogeneous Networks—A Deep Reinforcement Learning Approach

Abstract

In 5G scenarios, network slicing and multi-tier heterogeneous networks are critical to guarantee the service level agreement (SLA) of various services. In this letter, a dynamic radio resource slicing framework considering joint bandwidth slicing ratios and base station (BS)-user association is presented for a two-tier heterogeneous wireless network. This framework maximizes the spectrum reuse ratio through a two-step deep reinforcement learning (DRL) method, and guarantees the SLA of network slices simultaneously. Specially, a distributed agent (D-Agent) is deployed at each BS for acquiring the slicing resource in a single BS level. Meanwhile, a centralized agent (C-Agent) manages radio resource allocation and user association among heterogeneous BSs to guarantee the SLA.