An intelligent approach for classification of GPS satellites based on neural network, genetic algorithm and particle swarm optimization

Abstract

Geometric dilution of precision (GDOP) factor, which is broadly utilized in satellite navigation, denotes the additional multiplicative impact of navigation satellite geometry on positional measurement precision. This factor is frequently employed to select suitable satellites' subsets from at least 24 orbited existing satellites. The GDOP calculation has a time burden including complicated transformation and inversion of measurement matrices. To tackle this shortcoming, neural network- (NN-) based methods using the back propagation (BP) training algorithm have been broadly used. However, there are several parameters for the NN-based approaches that ought to be chosen by many trials. To alleviate this problem and enhance the BP training algorithm, we propose an intelligent approach based on the improved NN training methods and evolutionary algorithms (EAs), including namely, genetic algorithm (GA) and particle swarm optimization (PSO), to classify global positioning system (GPS) satellites using the GDOP factor. The simulation results for the real GPS GDOP data indicate that both the GA and PSO enhance the classification ratios, although the GA leads to higher ratios. The highest classification ratio is obtained by Levenberg-Marquardt training algorithm with GA.