TD3-based Stochastic Workload Offloading for 5G-based Cloud Control System

Abstract

The cloud control system is an emerging trend combining communication, computing, and automation techniques, it supports virtualization deployment of multiple programmable logic controllers (PLCs) in a central cloud server and interacts with remote actuators through communication networks. In the factory, there are massive control workloads in a workshop, which is usually managed by a cloud control system, it's an issue about how to make full use of communication and computation resources to bear as many control workloads as possible in this scenario. Therefore, a deep reinforcement learning (DRL) based stochastic workload offloading algorithm is studied in this paper, aiming to optimize the workload distribution and network resource allocation as well as to guarantee workload execution success rate. The simulation results demonstrate that the proposed DRL-based algorithm outperforms other benchmark ones.