Computation Offloading and Resource Allocation in Mobile Edge Computing-enabled IoT Network

Abstract

Mobile edge computation (MEC)-enabled Internet of Things (IoT) network have gained significant attention from academia and industry due to its ability to provide ultra-low latency computation services for several IoT applications such as VR/AR, smart city and online gaming, etc. This article explores the challenge problem of jointly addressing computation offloading and resource allocation (CORA) in a time-varying MEC-enabled IoT network. Firstly, we propose a multiuser and multiserver MEC-enabled IoT architecture based on centralized management for the decision-making process of CORA. Secondly, we formulate a task execution delay minimization problem considering energy consumption and resource constraints. It is challenging to solve this problem using traditional methods due to the coupling between the variables, as well as the dynamic nature of the network. To this end, we convert the original problem into a Markov Decision Process (MDP). Then, an entropy-based deep reinforcement learning algorithm (EDRL) with strong exploration capabilities and stability is used to learn the dynamic CORA strategy. Finally, extensive simulation experiments have indicated that our proposed EDRL method exhibits superior learning ability and stability compared to the DDPG. Except for the exhaustive method, EDRL outperforms other baselines with respect to execution delay and energy consumption.