Towards enhancing resource allocation and routing in network slicing for beyond 5g using deep reinforcement learning approach

Abstract

The future of global connectivity hinges on the successful development of beyond 5G networks. These next-generation networks promise to deliver seamless and intelligent experiences, but achieving this vision requires significant advancements in resource management, routing protocols and network security. This work explores these challenges and proposes solutions leveraging Deep Reinforcement Learning (DRL) and Software-Defined Networking (SDN) for a robust and secure 6G infrastructure. Beyond 5G applications, such as extended reality and connected vehicles, have diverse needs and require differentiated service levels to guarantee a high-quality user experience. Network slicing, a technique that virtualizes the physical network into isolated slices catering to specific applications, emerges as a potential solution. However, efficiently allocating resources within slices and dynamically adapting them to changing demands presents a complex challenge. This work proposes a DRL-based approach called customized sub-slicing to address this complexity. It advocates for splitting network slices (access, transport and core) into sub-slices, enabling granular resource allocation for various beyond 5G applications. newline