An Artificial Intelligence Radio Propagation Model Based on Geographical Information

Abstract

In this article, a new wireless network planning propagation model is proposed, the datasets are continuous wave data from different cities, and the model is constructed based on the artificial neural network in machine learning and uses geographical information to do feature engineering. Six geographical information features are used as the input feature vectors of the model. The terrain type near the receiving point is a corrective factor affecting the propagation field intensity. To make the physical implications more explicit, two multilayer perceptron networks were designed, and the one-hot encoding of terrain types of five grids near the receiving point on the line between the receiving and transmitting points is taken as one of the input features of the second network. Through the model training and validation, and tested in actual scene, the model on the datasets of different cities has achieved good results. Compared with the traditional propagation model, such as the standard propagation model (SPM), this artificial intelligence model is more suitable for correction of data containing random disturbance and has a better simulation accuracy; the random disturbance includes external random interference sources, global positioning system offsets, and inaccurate maps.