Abstract: Abstract: In view of the increasing demand for user experience quality in Vehicle Edge Computing (VEC), the difficulty in obtaining link status caused by highly mobile vehicles, and the time-varying nature of heterogeneous edge nodes providing resources to vehicles, a vehicular edge computing scheme with task offloading and resource optimization was developed. Firstly, a vehicle-to-infrastructure transmission model was proposed by comprehensively considering the intra-edge and inter-edge interference. By introducing non-orthogonal multiple access technology, edge nodes were not only not required to rely on link status information but also improved channel capacity. Secondly, in order to enhance the performance and efficiency of the system, a multi-agent dual-delay deep deterministic algorithm was designed to formulate task offloading strategies, which could be dynamically adjusted through interactive learning with the environment. Thirdly, the synergy of the two strategies was jointly considered, and an optimization scheme was formulated with the goal of maximizing the task service ratio to meet the increasing user experience quality requirements. Finally, simulation experiments were carried out using a real vehicle trajectory dataset. The results show that compared with the three current representative schemes (i.e., the schemes using Random Offloading (RO) algorithm, D4PG (Distributed Distributional Deep Deterministic Policy Gradient) algorithm and MADDPG (Multi-agent Deep Deterministic Policy Gradient) algorithm as task offloading algorithms), the average service ratio index of the proposed scheme was improved by more than 20%, 10% and 29% respectively in three scenarios (i.e., normal scenarios, task-intensive scenarios and delay-sensitive scenarios), verifying the advantages and effectiveness of the scheme.

Key words: Keywords: vehicular edge computing, non-orthogonal multiple access, deep reinforcement learning, task offloading, resource allocation

摘要： 摘 要: 针对车辆边缘计算（VEC）中存在的用户体验质量需求不断增加、高度移动车辆引起的链路状态获取困难、异构边缘节点为车辆提供资源的时变性等问题，制定了一种联合任务卸载和资源优化的车辆边缘计算方案。首先，在不失一般性的前提下，综合考虑边缘内和边缘间干扰，提出一种车辆到基础设施的传输模型，通过引入非正交多址接入技术使得边缘节点不仅无需依赖链路状态信息，还可以提升信道容量；其次，为增强系统的性能和效率，设计出一种多智能体双延迟深度确定性算法用于制定任务卸载策略，可通过与环境的交互学习动态调整任务卸载策略；再次，联合考虑两种策略的协同作用并制定优化方案，将最大化任务服务比率作为目标进行求解以满足不断提升的用户体验质量需求。最后，通过真实车辆轨迹数据集进行仿真实验。结果表明，与当前具有代表性的3种方案（即分别以随机卸载（RO）算法、D4PG（Distributed Distributional Deep Deterministic Policy Gradient）算法和MADDPG（Multi-agent Deep Deterministic Policy Gradient）算法为任务卸载算法的方案）在3类场景下（即普通场景、任务密集型场景和时延敏感型场景）相比，本文所提方案的平均服务比率指标分别提高了20%，10%和29%以上，验证了该方案的优势和有效性。

关键词: 关键词: 车辆边缘计算, 非正交多址接入, 深度强化学习, 任务卸载, 资源分配