GNSS receiver implementation using CUDA

Abstract

In this project we made the implementation of CUDA technology in a GNSS SDR in order to use parallel computing capabilities to solve a performance problem in GNSS signal processing, which were long execution times due to high computation demand of determinate tasks. Typically, execution times among the operations were very large and iterative due to complex algorithm usage and huge amounts of data manipulation, for this reason the aim of this research was to provide CUDA-MEX embedded files in order to reduce time consumption inside SoftGNSS blocks. The main block that has been optimized is the acquisition block (convolutions, correlations, transform, etc.) and we provided some resources to work with respective Tracking and Positioning blocks in future research. The novelty is that there are not CUDA-MEX embedded files destined to work with GNSS SDR. These MEX-CUDA files received real GPS raw signal (11ms) in order to make respective calculations to determine if PRN was found in a sampled signal. Most analysis