CNN-Based Propagation Loss Modeling Based on a Series of Images in Urban Scenarios

Abstract

Propagation loss modeling is considered to be important for the efficient planning of any wireless communication system. We present a novel convolutional neural network (CNN)-based propagation loss modeling based on field measurements in urban scenarios. We consider bird’s-eye images as input data that include the information of buildings, intersections, and roadways. Each image represents a different spatial segment of main propagation paths between a transmitter and a receiver. Multiple feature extraction layers based on CNNs are used. The proposed model shows its superior performance compared with the COST 231 Walfisch-Ikegami model.