Task Offloading Optimization in Vehicular Edge Computing with Meta-Policy Models

With the rapid development of mobile communication and Internet technology, the Internet of Vehicle (IoV) is becoming a key development direction in the future transportation field. Vehicle applications are generating more and more data. Offloading computational tasks to traditional cloud platforms often results in high task latency. To address these challenges, Vehicular Edge Computing (VEC) has become a promising solution, enabling the execution of complex computational tasks on edge servers deployed within roadside units (RSUs). By reducing data transmission delays and providing localized computing capabilities, VEC significantly enhances the performance of IoV systems. In this paper, we propose an intelligent task offloading algorithm based on a meta-reinforcement learning framework. Our approach models vehicle applications as directed acyclic graphs (DAGs), representing intricate dependencies among computational tasks. Using recurrent neural networks (RNNs), our algorithm learns optimized offloading strategies through training, dynamically adapting to varying task demands and network conditions. The experimental results indicate that the vehicle task offloading strategy obtained using this method can achieve low energy consumption while ensuring low latency. And this method leverages the computational capabilities of edge servers to enable efficient resource allocation and optimized scheduling.