Advanced GNSS-R Signals Processing With GPUs

Abstract

Global navigation satellite system reflectometry (GNSS-R) is a group of techniques that uses satellite navigation signals as signals of opportunity for remote sensing applications. In GNSS-R, large amounts of data are acquired and have to be processed. Computation time is typically the bottleneck for ground and airborne experiments. This article presents an efficient solution for off-line GNSS-R processing data taking advantage of graphics processing units (GPUs). After comparing to the typically used CPU languages, such as MATLAB and C++, the advantage of using parallel processing on the GPU is clear. GPU-based computation can reduce the processing time by as much as 95% of the acquisition time of the data. An implementation taking advantage of a home-use GPU is proposed for the data processing units. Thanks to its efficiency, even real-time processing experiments are feasible.

Keywords

• Compute unified device architecture (CUDA)
• global navigation satellite system reflectometry (GNSS-R)
• graphics processing unit (GPU) processing
• parallel computing
• real-time processing