Integration of Artificial Neural Network and the Optimal GNSS Satellites’ Configuration for Improving GNSS Positioning Techniques (A Case Study in Egypt)

Nowadays, the global navigation satellite system (GNSS) positioning techniques based on the International GNSS Service (IGS) products are extensively used for various precise applications. However, specific conditions such as the dual-frequency observations and the final IGS products are required. Consequently, the absence of the final IGS data and using single-frequency observations will degrade these techniques’ accuracy. In this paper, two algorithms through two separated stages are formulated for improving the single-frequency GNSS observations by using one GNSS receiver based on the broadcast ephemerides in real time or close to real time. The first algorithm represents the preparation stage for the second one. It classifies the observations by separating the optimal values of position dilution of precision (PDOP) and the number of satellites (NOS), as well as the corresponding values of coordinates. The second stage includes an algorithm based on the artificial neural network (ANN) approach, which is set at the ANN variables that produce the best precision through the applied tests at the present study. Binary numbers, log sigmoid-Purelin, cascade forward net, and one hidden layer with a size of 10 neurons are the optimal variables of ANN inputs format, transfer functions constellations, feedforward net type, and the number of hidden layers (NHL) and its size, respectively. The simulation results show that the designed algorithms produce a significant improvement in the horizontal and vertical components. Lastly, an evaluation stage is performed in the case of dual-frequency observations by using broadcast ephemerides. The simulation outputs indicate that the precision at applying the proposed integration is completely enhanced compared with the outputs of IGS final data.