SOIS-A2C Scheme: Facilitating Management of Multi-Radio Resources in Heterogeneous Inter-RAN Slicing in the Presence of Hardware Impairments

Recent years have witnessed the emergence of the concept of network slicing (NS) that enables the creation of independent, virtualized logical networks on the same physical infrastructure. Each NS is tailored to meet the needs of a particular service or application. NS is widely considered a key enabling technology for end-to-end (E2E) automation in managing resources within radio access networks (RAN). To achieve E2E automation in RAN slicing, it is essential to automate resource allocation at both the intra- and inter-slice levels to meet the demands of future applications and services. In this context, the present study focuses on the automated management of multiple radio resources at the inter-slice level. More specifically, we propose a self-optimizing inter-slice scheme based on the deep reinforcement learning (DRL) advantage actor-critic (A2C) algorithm, named SOIS-A2C. Our goal is to maximize the spectral efficiency of the system while maintaining high service quality by considering the effects of hardware distortions and intra-slicing interference. The results highlight the effectiveness of the proposed SOIS-A2C scheme, which demonstrates superior performance in maximizing spectral efficiency as compared to benchmark schemes such as the SOIS-based deep Q-network (DQN) and hard slicing in highly fluctuating environments and under ideal and non-ideal hardware conditions.