Neural Network Prediction of Receiver Bias in Ionospheric Delay Computation

Abstract

An important measure typically used to understand ionosphere properties and disturbances is total electron content (TEC). A typical approach to calculating the ionospheric TEC is by analyzing dual-frequency GPS data. Satellite and receiver biases are the primary discrepancies in TEC computation. In this work, we develop a neural network to predict the instrumental receiver bias based on slant TEC. The minimum standard deviation method is used to calculate the receiver bias. Neural network with two hidden layers is trained with datasets and then used to predict the receiver bias. The predicted receiver bias from the proposed neural network differs from the baseline method by about 10 to 20 percent.