CNN Distance Estimation Based on Received Signal Strength Indicator

With the advancement of technology, an increasing number of applications rely on positioning services. While GPS technology is well-developed for outdoor environments, providing accurate location information indoors remains crucial. Distance estimation is vital component of indoor positioning. However, the typical path loss model (PLM) has limitation in complex indoor settings. This article proposes using deep learning methods to enhance distance estimation. By collecting Wi-Fi signals form reference points and utilizing a 1D Convolutional Neural Network (lDCNN) model for distance classification, we can reduce outliers. In addition, as the number of classification categories increases moderately, the distance estimation error can be effectively reduced. Experimental results show the when the distance classification categories are set to 3, 6 and 12 respectively, the average distance estimation errors are 0.379 meters, 0.113 meters and 0.058 meters.