FURI | Summer 2020

Leveraging on Deep Learning to Predict the Optimal Beam Index Using Wireless Sensing Localization

Data icon, disabled. Four grey bars arranged like a vertical bar chart.

The demand for communication networks with high capacity rate will be fulfilled with the recent developments of 5G technology. But then this technology utilizes mmWave systems, which has the following challenges; sensitivity to blockage and the large beam training overhead in large antenna array. This project turns to leverage machine learning to predict the optimal beam direction and Line-of-Sight (LOS) connection at no training overhead, using real measurements from a testbed. The dataset from this testbed is exploited to build a simple multilayer neural network, capable of predicting the optimal beam direction of the mmWave at a considerable success probability using only position information of the transmitter and receiver.

Student researcher

Tawfik Mohammed Osman

Electrical engineering

Hometown: Tamale, Northern Region, Ghana

Graduation date: Fall 2020