An artificial neural network approch for wireless sensor networks localization with RSSI value

A great deal of method, algorithm and study on distance calibration and positioning through RSSI in wireless sensor networks are available. In this dissertation, it is intended to provide a different perspective using artificial neural networks to distance calibration and positioning through RSSI. To that end, a mobile tool carrying wireless sensor node has been designed. In an experiment environment where this tool has a wireless sensor node at its each four corners and each unit equals to 60 cm at 16x14, the distance to corner nodes with 60 cm advancement – RSSI values and x, y coordinates of the tool have been recorded in the computer. In order to mitigate the effects of measurements affected by the noise on the results, these processes have been repeated 100 times for each node at each point and consequently, a total of 4x100x(16x14) RSSI distance and positioning information have been recorded. In the next stage, these data have been used as training and test data in artificial neural networks designed in MATLAB environment and the functions of artificial neural networks, whose weight was counted, have been obtained. Similarly again in MATLAB environment, a function was coded for Bounding Box algorithm and both results obtained from Bounding Box algorithm and ANN's have been compared. While realizing the application, in addition to Arduino IDE, MATLAB, Netbeans IDE software realization environments, MySql database has been used in order to store the data obtained from the application. Wireless sensor nodes have been produced using Arduino and XBee modules and the configuration of XBee modules has been made through USB using XCTU interface.