A Dueling DQN-Based Computational Offloading Method in MEC-Enabled IIoT Network

Abstract Industrial Internet of Things (IIoT) is a promising mechanism of Industry 4.0. Mobile edge computing (MEC) is an emerging mechanism that is an enabler for IoT applications, which applies computation offloading mechanism to offload workload from IoT devices to MEC servers to enhance computing quality. This paper addresses the computation offloading problem under a heterogeneously loaded MEC-enabled IIoT network, which specifically applies one-hop offloading for devices and task queue for devices and servers. We first formulate our computation offloading problem as a Markov Decision Process, then design a dueling deep Q-network-based computation offloading method for better optimizing the value function for specific state and advantage function for specific action. Experimental results verify that our scheme reduces more energy consumption (30%), latency (50%) and task dropped rate (50%) of IIoT devices, compared with other popular methods.