Improvement RTDGPS Accuracy and Stability by using RNN and PSO Hybrid Model

Stability of the "Real Time Differential Global Positioning System (RTDGPS)" in servicing the users is of much significance. This system consists of two "reference and mobile stations Interruption of the signal of reference station leads to reduce system stability. This issue results in the lack of production of system specific protocol in the reference system, thus the available corrective factors in this protocol, would not reach to the receiving user through the linking interface. In this case, the system enters the DGPS state from GPS, and the accuracy of positioning system is declined. In this paper, in order to compensate the lack of signal, the model "Recurrent Neural Network (RNN) prediction" has been used. Training the network was done through the Particles Swarm Optimization (PSO). The correction pseudo-distance of each satellite is the input of predictor model. PSO-RNN experiment using the real data Pseudo Range Correction (PRC) indicated that the RMS error of prediction model was 0.18m. The reference station and the user of system were implemented using the cheap receivers. The designed interface software performs the operations of reference station. The accuracy of positioning of the system's user station was obtained 0.4 in static state through adding the model PSO-RNN to the reference station, using the practical experiments. This amount of accuracy demonstrates improvement rather to the state, which the PSO-RNN is not utilized in the reference station.