Convolutional Neural Network-based UWB System Localization

Abstract

In this paper, we introduce a localization method that employs the ultra-wideband (UWB) technology and a convolutional neural network (CNN). The CNN model is designed to estimate the location of the device which transmits the UWB signals to three different receiver antennas under an indoor environment. The proposed method uses the red, green, and blue (RGB) image that is generated from the received signals. The location of the transmitter is directly estimated from the RGB image through the designed CNN without predicting the distance between each receiver and transmitter. The simulation results show that our proposed CNN-based localization method has lower root-mean-square-error by approximately 0.5 meters than the previous CNN-based method. As a square map size grows, the performance improvement increases. These results verify the significant improvement in the performance of our proposed CNN-based localization method compared to the time of arrival method and previous CNN-based method.