Application of artificial neural networks for path loss prediction in railway environments

Abstract

To balance the precision and generality of the prediction model, a new path loss artificial neural network (ANN) prediction model for railway environments is presented firstly in this paper. The utilization of back propagation ANN can overcome some disadvantages of such conventional prediction models as empirical and deterministic models. The training data is based on the electric field strength measurements in the Zhengzhou-Xi'an express railway line in China. Through many attempts and comparisons, the suitable architecture and learning algorithm are chosen in the proposed model. After training, the proposed model can predict the path losses accurately in typical similar railway environments. Comparisons between a conventional model and the proposed model, with the measured and predicted data show that the proposed model is sufficiently applicable in railway scenarios.