Parallel resolution of large-scale GNSS network un-difference ambiguity

Abstract

To increase redundant observations and estimate fractional cycle bias (FCB) in a global network and whole session, hundreds of globally distributed Global Navigation Satellite System (GNSS) tracking stations are required in the server side. However, the improvement of computational efficiency for FCB estimation and un-difference ambiguity fixing is a critical issue. In this paper, using a multi-node and multi-core platform based on Task Parallel Library, a strategy for multi-core un-difference parallel resolution is proposed. Based on MapReduce, a workflow for multi-node FCB parallel estimation and un-difference ambiguity parallel fixing is developed. As a result, the efficiency of FCB estimating and ambiguity fixing is improved significantly. Data from global International GNSS Service (IGS) tracking stations are used in the experiment. In the server side, the speed-up ratio of FCB estimation using a six-node and four-core platform reaches 14.76 times. In the user side, globally distributed user stations are applied to parallel ambiguity fixing, whereupon the speed-up ratio under the same platform is improved to 12.33 times. In addition, the average accuracy of static hourly solutions for 16 user stations improves from 2.50, 3.12 and 0.99cm to 1.30, 0.77 and 0.65cm in the vertical, east and north components, respectively.