【摘要】 为了解决传统移动边缘计算网络无法很好地支持车辆的高速移动性和动态网络拓扑,设计了车辆多址接入边缘计算网络,实现路边单元和智能车辆的协同计算迁移。在该网络架构下,提出了多址接入模式选择和任务分配的联合优化问题,旨在最大化系统的长期收益,同时满足多样化的车联网应用需求,兼顾系统的能量消耗。针对该复杂的联合优化问题,设计了基于深度增强学习的多址接入协同计算迁移策略,该策略能够很好地克服传统Q-learning算法因网络规模增加带来的维度灾难挑战。仿真结果验证了所提算法具有良好的计算性能。更多还原

【Abstract】 In order to solve the problem that traditional mobile edge computing network can’t be straightforwardly applied to the Internet of vehicles(IoV) due to high speed mobility and dynamic network topology, a vehicular edge multi-access computing network(VE-MACN) was introduced to realize collaborative computing offloading between roadside units and smart vehicles. In this context, the collaborative computation offloading was formulated as a joint multi-access model selection and task assignment problem to realize the good balance between long-term system utility, diverse needs of Io V applications and energy consumption. Considering the complex joint optimization problem, a deep reinforcement learning-based collaborative computing offloading scheme was designed to overcome the curse of dimensionality for Q-learning algorithm. The simulation results demonstrate that the feasibility and effectiveness of proposed offloading scheme.更多还原