LTE Network Planning Using AI and AR

In this paper, an optimal base station placement for a LTE cellular network is provided. The designed system relies on a web server that communicates with pre-trained machine learning models trained using measured data and deployed on MATLAB For training, a fingerprinting database has been created for a target RSSI versus the GPS coordinates as features in an outdoor and indoor environment on a multi floor buildings in the university campus. Gaussian Process Regression (GPR) has been used as a benchmark machine learning algorithm due to its outstanding performance. However, its performance was compared with different machine learning techniques in terms of computational complexity and accuracy. Validation of the system has shown that GPR outperformed the other techniques with mean square error of 4.3% at the expense of the time required for training. Moreover, the application has been provided with an augmented reality interface that shows the placed BS on the map along with the predicted RSSI value and the GPS coordinates. The whole system is designed on unity software and made available to mobile devices.