Joint deep reinforcement learning strategy in MEC for smart internet of vehicles edge computing networks

Abstract

The Internet of Vehicles (IoV) has a limited computing capacity, making processing computation tasks challenging. These vehicular services are updated through communication and computing platforms. Edge computing is deployed closest to the terminals to extend the cloud computing facilities. However, the limitation of the vehicular edge nodes, satisfying the Quality of Experience (QoE) is the challenge. This paper developed an imaginative IoV scenario supported by mobile edge computing (MEC) by constructing collaborative processes such as task offloading decisions and resource allocation in various roadside units (RSU) environments that cover multiple vehicles. After that, Deep reinforcement Learning (DRL) is employed to solve the joint optimisation issue. Based on this joint optimisation model, the offloading decisions and resource allocations are gained to reduce the cost obtained in end-to-end delay and expense of resource computation. This problem is formulated based on the Markov Decision Process (MDP) designed functions like state, action, and reward. The proposed model's performance evaluations and numerical results achieve less average delay for 30 vehicle nodes in simulation.