Deep Reinforcement Learning and Nash Equilibrium Game-Based Task Offloading Optimization Strategy for the IoV

With the rapid development of 5G, the amount of data generated by the Internet of Vehicles (IoV) is growing explosively, which affects the efficiency of vehicle task offloading. To solve this problem, this paper proposes a task offloading strategy based on Regional Roadside Unit Fusion (RRF) algorithm. Firstly, with the help of sensors to collect vehicle historical trajectory data as input, the Deep Reinforcement Learning (DRL) technique combined with Convolutional Neural Network (CNN) and path planning algorithm is used to accurately project the dynamic behavior of vehicles. Secondly, when a Roadside Unit Fusion (RSU) is unable to handle a task, the Automatic Tagging (AT) algorithm will tag the RSU, preparing for the optimization of subsequent RSU. Finally, the Nash Equilibrium Game (NEG) algorithm is used to optimize RSU resource allocation to maximize task offloading efficiency and reduce energy consumption. Experimental results show that the RRF algorithm proposed in this paper has significant advantages over the standard Full DRL algorithm and the Non-Orthogonal Multiple Access (NOMA) algorithm. In terms of task offloading delay, it reduces 11.2% and 9.6%, respectively, energy consumption is reduced by 12.7% and 8.2%, respectively, and the task offloading success rate is improved to 92%, which is significantly better than the comparison algorithms. The algorithm provides a better solution for the practical application of Mobile Edge Computing (MEC) in the IoV.