Collective reinforcement learning based resource allocation for digital twin service in 6G networks

Abstract

The 6th generation (6G) mobile communications technology will realize the interconnection of humans, machines, things as well as virtual space. The development of digital twins (DTs) and 6G has accelerated the Internet of Things (IoT) in an unprecedented way. The combination of DTs and edge intelligence (EI) enables powerful digital space synchronized with the real world constructed in the intelligent edge, bringing real-time, and adaptive services delivery of IoT. However, the dynamic features and heterogeneous resources in 6G-enabled IoT make the resource allocation for computation-intensive and delay-sensitive DTs services more challenging. In this paper, we first define the DTs implementation process as a DT service function chain (DTSFC) and address the resource allocation problem of DTs-empowered networks in form of dynamic DTSFCs orchestration. We further propose a novel collective reinforcement learning (CRL) method which is inspired by human collaboration, to realize the effective resource allocation of DTSFCs. Numerical results verify that the proposed CRL algorithm improves the learning efficiency and generalization ability compared with the benchmarks.