Slice admission control in 5G wireless communication with multi-dimensional state space and distributed action space: A sequential twin actor-critic approach

Abstract

Network slicing represents a paradigm shift in the way resources are allocated for different 5G network functions through network function virtualization. This innovation aims to facilitate logical resource allocation, accommodating the anticipated surge in network resource requirements. This will harness automatic processing, scheduling, and orchestration for efficient management. To overcome the challenge of managing network resources under heavy demand, slice providers need to leverage both artificial intelligence and slice admission control strategies. While 5G network resources can be allocated to maintain a slice, the logical allocation and real-time network evaluation must be continuously examined and adjusted if network resilience is to be maintained. The complex task of leveraging slice admission control to maintain 5G network resilience has not been fully investigated. To tackle this problem, we propose a machine learning approach for slice admission control and resource allocation optimization so as to maintain network resilience. Machine learning algorithms offer a powerful tool for making robust and autonomous decisions, which are crucial for effective slice admission control. By intelligently allocating resources based on real-time demand and network conditions, these algorithms can help ensure long-term network resilience and achieve key objectives. While various machine learning algorithms hold promise for 5G resource management and admission control, reinforcement learning (RL) has emerged as a particularly exciting solution. Its ability to mimic human learning processes makes it a versatile solution, well-suited to tackle the complex challenges of network control. To fill this gap, we propose a new technique known as sequential twin actor critic (STAC). Simulations show that the STAC improves network resilience through enhanced admission probability and overall utility.