Improving DGPS Accuracy Using Predictions of Reference Position Components Error Factors

For precise locating, Differentials Global Positioning System requires prediction of differential corrections for the future times. The system is comprised of both fixed and mobile stations. If the satellites of the two stations are exactly the same, the sources of errors will be close to each other at the two stations; in this case, reference position components factors can be used as corrective factors for offsetting user station positioning error. In this paper, Genetic and Artificial Neural Network hybrid algorithms (Evolutionary Neural Network), Support Vector Machines, Autoregressive Moving Average and Recurrent Neural Network have been used for corrections. In order to test the algorithms, static sampling of the position data of an inexpensive receiver was used and the predicted reference position components error corrections were applied elsewhere. The tests performed as post-process showed that the positioning RMS error decreases up to 0.5 m. The evolutionary neural network prediction model is more accurate than other models and its RMS error is 0.12 m.