Abstract: In the last-mile delivery scenario of Unmanned Aerial Vehicles (UAVs),the existing cloud computing architecture has a high latency problem,which cannot meet the execution requirements of artificial intelligence applications.The edge computing architecture can meet the needs of artificial intelligence applications by sinking computing resources to the edge with its low latency and high computing capabilities.However,most of the current researches are limited to a single edge server and lack the design of a parallel collaboration framework.According to the characteristics of MEC environment and the last mile of UAV delivery process,the computing load of edge server was fully considered and a multi-edge server parallel task processing framework was designed based on collaborative device and edge.The scheduling algorithm of Shortest Scheduling Latency First (SSLF) task was improved and α-SSLF algorithm was designed.The α-SSLF algorithm could fully consider optimizing task execution time under unstable network actual data rate.Experimental result showed that the proposed framework was superior to the traditional serial task processing framework in processing latency.

Key words: unmanned aerial vehicle delivery, mobile edge computing, task scheduling, collaborative device and edge, parallel computing

摘要： 在无人机最后一公里配送场景中,现有的云计算架构存在高时延问题,无法满足人工智能应用的执行需求。边缘计算架构通过将计算资源下沉到边缘,以其低时延、高计算能力的特点,可以满足人工智能应用的需求。但是目前的研究大多局限于单个边缘服务器,缺少并行协同框架的设计。为了解决该问题,本文首先根据移动边缘计算环境和无人机最后一公里配送过程的特点,充分考虑边缘服务器的计算负载问题,设计了基于端边协同的多边缘服务器并行任务处理框架;然后在该框架上对最短响应时间优先的任务调度算法进行改进,设计了α-SSLF算法。该算法能够考虑在网络实时数据率不稳定的情况下,充分优化任务执行时间。结果表明,基于端边协同的多边缘服务器并行任务处理框架在处理时延上优于传统的串行任务处理框架。

关键词: 无人机配送, 移动边缘计算, 任务调度, 端边协同, 并行计算