雷达预警系统多目标轨迹预测的并行处理方法  

Parallel Processing Method of Multiple Targets Trajectory Prediction in Radar Early Warning System

作　　者：王雪[1] 施岩龙[2] 袁家斌[1] 

机构地区：[1]南京航空航天大学,南京210016 [2]南京电子技术研究所,南京210013

出　　处：《小型微型计算机系统》2016年第4期835-840,共6页Journal of Chinese Computer Systems

基　　金：江苏省普通高校研究生科研创新计划项目(SJLX_0138)资助;国家自然科学基金重点项目(61139002)资助

摘　　要：多目标实时轨迹预测是空中目标预警的核心任务之一.随着雷达(组网)监测范围的扩大,数据量激增,现有的串行处理方式已不能满足实时性要求.本文针对多目标实时轨迹预测中对同一批目标定轨及轨迹修正需进行数据回溯的特点,采用目标任务与计算节点绑定的方法,设计出一套完整的分布式异构并行方案,解决了通常的分布式雷达数据处理系统在处理该任务时因任务与计算节点相互独立导致的通信量大的问题.方案包括系统架构、进程设计等,并根据轨迹预测任务并行化设计特点进行建模,提出两种任务调度策略,在分布式集群上加以实现,结果表明该方案获得了良好的性能.The real time trajectory prediction of multiple targets is one of the core tasks in the early warning system for air targets.While radar＇s monitoring scope expands,the amount of the data to process soars and the existing approaches based on serial processing can hardly process so much data in real-time.Ordinary distributed radar data processing systems have to transfer tremendous amount of data because of the separation of the tasks and the computing nodes.Given the fact that data backtracking is necessary for orbit determination and trajectory correcting of the same target,we propose a distributed heterogeneous parallel solution by binding tasks and computing nodes to reduce the data to transfer.The solution contains system architecture and our design of threads.We also model the parallel design of trajectory prediction tasks and put forward two task scheduling strategies.The experiments on a distributed cluster showthat out solution improves the performance greatly.

关 键 词：轨迹预测 分布式并行处理 调度策略 实时性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]