Localization by using Deep Neural Networks for Long Range Target

In this paper, deep neural networks are proposed as a new and original approach for localization problem. For the proposed method, a deep network was first created and trained for positional information generated by using directional information generated by systems that are located in different positions and consisting of a arrays of sensors. Then, test data were generated for different locations and the trained deep network was tested. The same test data are fed to the linear least squares algorithm, which is frequently used for the localization problem in the literature, but the position accuracy decreases as it moves away from the target due to the spherical shape of the earth for long-range targets. The methods were subjected to repeated tests for locations at different distances and it was observed that the deep neural network method gave more accurate results compared to the traditional method as the distance of the source of the broadcast to be located from the sensors increased.