Applying Neural Network Ensembles for Clustering of GPS Satellites

  • Mosavi M. R.

要旨

Several external sources introduce errors into a GPS position estimated by a GPS receiver. One important factor in determining positional accuracy is the constellation, or geometry, of the group of satellites from which signals are being received. Geometric Dilution of Precision (GDOP) allows to determine the good selected satellite geometry and to give an approach about the GPS accuracy. The most correct solution to get GPS GDOP is to use inverse matrix on all the combinations and selecting the lowest one, but inversing a matrix puts a lot of computational burden on the navigator’s processor. In this paper, the clustering of GPS satellites is done based on the GDOP resulting from the Neural Network (NN), NN ensemble and NN ensemble by using Genetic Algorithm (GA) to select the appropriate subset of navigation satellites. The proposed methods are simulated and validated by a software simulation. These algorithms provide a realistic computational approach without needing to calculate the inverse matrix. The simulation results demonstrate that NN ensemble with GA has greater accuracy and can improve clustering accuracy of GPS satellites about 99%.
出版済
2011-09-01
セクション
Article