摘要
基于精密单点定位技术的非差模式是当前GNSS数据处理的主要策略之一。随着测站规模的增大,非差模式的处理时间也线性递增,传统的串行处理方法需消耗大量的计算时间。采用工厂模式和责任链模式实现了非差精密单点定位;利用轻量级的并行编程技术从底层设计并实现了基于任务的非差多核并行解算;进一步在网络多节点环境中建立并发布非差计算服务,实现了网络多节点协同并行解算GNSS数据。通过大量数据的测试与试验,验证了多核多节点的非差并行解算方案的高效性。试验结果表明,单节点多核并行、双节点网络并行、四节点网络并行、六节点网络并行的计算效率分别比单节点串行方案平均提高了2.74,5.30,9.38和14.69倍。
The undifferenced pattern based on precise point positioning（ PPP） technology is one of the main strategies of the current Global Navigation Satellite System（ GNSS） data processing. With the increasing scale of the stations, the processing time of undifferenced model is also linear increasing, and the traditional serial processing methods need to consume a lot of computing time. The PPP technology is implemented by using the factory design pattern and the chain of responsibility design pattern, and the parallel computing of PPP is designed from the bottom. In order to realize the network multi-node co-processing GNSS data, a lightweight multi-core undifferenced parallel computing based on the task is realized by adopting task parallel extensions technology, and the undifferenced computing services are established and released on the network nodes by using windows communication foundation technology. The high efficiency of undifferenced parallel computing scheme under multi-core multi-node platform is validated through a large amounts of data experiments, and the results show that compared to the singlenode serial processing, the average increase rate of computational efficiency is 2.74 in single-node multi-core parallel computing, 5.30 in two-node network parallel computing, 9.38 in four-node network parallel computing, and14.69 in six-node network parallel computing.
出处
《测绘科学技术学报》
CSCD
北大核心
2015年第6期565-569,共5页
Journal of Geomatics Science and Technology
基金
国家自然科学基金项目(41274015)
国家863计划项目(2013AA122501)
地理信息工程国家重点实验室开放研究基金项目(SKLGIE2014-M-1-5)
关键词
多核多节点
非差模式
大型GNSS网
并行计算
工厂设计模式
任务并行库
multi-core multi-node
undifferenced model
large GNSS network
parallel computing
factory design pattern
task parallel library
