Summary
International Conference on Emerging Technologies for Communications
2022
Session Number:O3
Session:
Number:O3-2
Machine learning method for location estimation at various altitudes using multiple items of sensed information in indoor environment
Ren Kawamura, Eisuke Kudoh,
pp.-
Publication Date:2022/11/29
Online ISSN:2188-5079
Summary:
In an indoor environment, it is difficult to receive radio waves directly from satellites which hinders accurate location estimation by satellite signals. Meanwhile, mobile communication propagation channels suffer from fading and shadowing, so estimation of indoor locations by using only the received signal power of a radio wave is inaccurate as well. Sensed information (e.g., temperature, humidity, illuminance) is often location dependent, and a location can be estimated accurately if such information is used in addition to the received signal power. The use of machine learning for an optimization algorithm has been shown to be promising. In this paper, we apply machine learning for indoor location estimation at various altitudes using multiple items of sensed information. We propose two types of neural networks, a two-dimensional neural network and a nearest node neural network, and experimentally evaluate them in an actual building. The results indicate that location estimation using the nearest node neural network has a greater coincidence probability than that using the two-dimensional neural network.