Optimal Task Offloading Strategy for Vehicular Networks in Mixed Coverage Scenarios
Xuewen He, Yuhao Cen, Yuhuan Liao, Xin Chen, Chao Yang · Applied Sciences · 2024
With the rapidly escalating demand for high real-time performance and data throughput capabilities, the limitations of on-board computing resources have rendered traditional computing services inadequate to meet these burgeoning requirements. Vehicular edge computing offers a viable solution to this challenge, yet the roadside units (RSUs) are prone to overloading in congested traffic conditions. In this paper, we introduce an optimal task offloading strategy under congested conditions, which is facilitated by a mixed coverage scenario with both 5G base stations and RSUs with the aim of enhancing the efficiency of computing resource utilization and reducing the task processing delay. This study employs long short-term memory networks to predict the loading status of base stations. Then, based on the prediction results, we propose an optimized task offloading strategy using the proximal policy optimization algorithm. The main constraint is that the data transmission rates of users should satisfy the quality of service. It effectively alleviates the overload issue of RSUs during congested conditions and improves service quality. The simulation results substantiate the effectiveness of the proposed strategy in reducing the task processing delay and enhancing the quality of service.