Wireless Indoor Localization Problem with Artificial Neural Network

Positioning in indoor application is a challenging problem with GPS signals. Because the obstacles such as doors and walls weaken the GPS signal amplitudes, indoor positioning results are not satisfying with the global positioning system. Indoor positioning may be critical for a variety of applications such as detecting the number of people, locating criminals in bounded regions, and obtaining the number of users in a special area. The Wi- Fi signal strength may be a key point to solve this problem. With several routers, the received Wi-Fi signal power information may use to determine the indoor localization using the information of routers location. In this work, Multi-Layer Perceptron (MLP) neural network method is proposed that can be implemented in monitoring and tracking devices. In the end, the theoretical background and simulation results are shared. Both k-fold cross-validation and hidden neuron numbers are changed in the simulation then the results are compared.

I. Introduction

Global Positioning System (GPS) is a very common application for localization. It is very effective in outdoor localization. There are several satellites in space to serve devices and receiving signals from more satellites improves position resolution. However, GPS indoor positioning has some limitations due to the signal strength affected by the doors, walls or as in underground places such as metro stations.