基于多集群架构的并行规划平台研究

发布时间：2018-03-28 03:26

本文选题：任务规划　切入点：集群　出处：《天津大学》2013年硕士论文

【摘要】：任务规划技术是随着无人飞行平台实际使用需求而迅速发展起来的一个新兴技术，它是以先进、高效的计算机平台为基础，通过对各种海量基础规划数据的计算、处理和分析，辅助制定无人飞行平台任务规划和最终飞行航迹。任务规划信息处理具有数据量大、计算复杂、处理时间长等特点，，基于多集群开发一个并行数据预处理系统对缩短任务规划时间具有重要意义。针对任务规划中地图预处理的计算密集和高度并行的特点，本文设计并实现了一个专用于调度地图预处理作业的作业调度系统。为验证作业调度系统的调度效率、容错性和规划平台的可扩展性，本文从任务规划中最耗时的基础地图数据的预处理入手，实现了地形适配区选择的串行计算，并运用OpenMP和MPI两种并行标准，实现了地形适配区选择的多线程并行计算和多节点并行计算。在集群平台上用作业调度系统对地形适配区选择作业进行调度，通过对比实验，表明了该作业调度系统具有良好的调度效率、容错性和规划平台良好的可扩展性，能充分利用集群的计算能力，从而缩短任务规划时间。
[Abstract]:Mission planning technology is a new technology developed rapidly with the actual demand of unmanned flight platform. It is based on advanced and efficient computer platform, through the calculation, processing and analysis of various mass basic planning data. To assist in developing mission planning and final flight path of unmanned flight platform. Mission planning information processing has the characteristics of large amount of data, complex calculation, long processing time, etc. It is important to develop a parallel data preprocessing system based on multiple clusters to shorten task planning time. According to the characteristics of intensive and highly parallel map preprocessing in task planning, this paper designs and implements a job scheduling system dedicated to scheduling map preprocessing, in order to verify the scheduling efficiency of the job scheduling system. Fault tolerance and extensibility of the planning platform. This paper starts with the preprocessing of the most time-consuming basic map data in task planning, realizes the serial calculation of terrain adaptation area selection, and uses two parallel standards, OpenMP and MPI. The multi-thread parallel computing and multi-node parallel computing for terrain adaptation area selection are realized. The job scheduling system is used to schedule the terrain adaptation area selection job on the cluster platform. It is shown that the job scheduling system has good scheduling efficiency, fault tolerance and good scalability of the planning platform, and can make full use of the computing power of the cluster, thus shortening the task planning time.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP338.6

【参考文献】