Utilizing hybrid recurrent neural network and genetic algorithm for predicting the pseudo-range correction factors to improve the accuracy of RTDGPS

Abstract

Positioning in real time with high precision is realized through implementing a Real-time Differential GPS (RTDGPS). Amccuracy of the system depends on sequential and uninterruptible transmission of Pseudo-Range Correction (PRC) codes from Reference Station (RS) to the user station. This transmission takes place in RTCM protocol. In order to compensate the signal absence and transmission delay of differential corrections, future values of the PRC should somehow be estimated. To this end a combination of Recurrent Neural Network (RNN) and Genetic Algorithm (GA) has been used. Addition of this algorithm not only improved the efficiency and consistency of the RTDGPS service, but also increased the accuracy of the positioning system. In this research RTDGPS system has been implemented using two inexpensive receivers in reference and user stations. Simulation and experimental results indicate that RTDGPS accuracy improved by predicting the future values of PRC. Time step of the prediction algorithm is considered to be 5 s. RMS value of the positioning error for a fixed point after addition of the RNN-GA algorithm to RTDGPS system has decreased.