面向边缘智能的车联网通信和计算资源联合管理策略

Joint Communication and Computing Resource Allocation for Edge Intelligence in Vehicular Networks

海量数据驱动未来车联网向智能化演进,计算密集型业务的激增给网络中通信和计算资源的管理带来了极大的挑战。为了解决上述问题,提出了面向边缘智能的车联网通信和计算资源联合管理策略,在考虑各边缘节点内存容量的前提下,通过模型切分将适量计算任务卸载到最优边缘节点,提高任务执行率并降低系统能耗。上述资源联合管理策略可建模为动态优化问题,传统的优化方法难以求解。因此,将人工智能技术应用到边缘计算领域,采用多智能体深度强化学习方法合理分配网络频谱资源和计算资源,提升网络性能。

Massive data drives the evolution of future vehicular networks to intelligence.With the explosion of computation-intensive services,the communication and computing resource allocation in future vehicular networks brings great challenges.To address this issue,a joint communication and computing resource allocation scheme for edge intelligence in vehicular networks is proposed.Under the premise of considering the memory capacity of each edge node,the computation task is offloaded to the optimal edge node via model segmentation,which improves the task execution efficiency and reduces the system energy consumption.The above joint resource allocation scheme can be modeled as a dynamic optimization problem,which is difficult to be solved with traditional optimization methods.Therefore,we apply artificial intelligence technology to edge computing,and adopt a multi-agent deep reinforcement learning method to rationally allocate spectrum resources and computing resources to improve network performance.