Publication Date

Spring 2023

Degree Type

Master's Project

Degree Name

Master of Science (MS)


Computer Science

First Advisor

Navrati Saxena

Second Advisor

Robert Chun

Third Advisor

Abhishek Roy


Beam-selection, Vision Transformer, mmWaves, vehicle- to-everything, LIDAR, GPS


5G networks explore mmWave technology to achieve faster data transfer and higher network capacity. The reduced coverage area of mmWaves creates the need to deploy large antenna arrays. However, beam sweeping across a large number of antenna arrays typically involves high overhead and latency. In a vehicle-to- everything (V2X) system, beam selection becomes a frequent process in the case when vehicles are moving at high speed, leading to frequent connection delays. Modern-day vehicular systems are integrated with advanced sensors like global positioning system

(GPS), light detection and ranging (LIDAR), radio detection and ranging (RADAR), etc. Machine learning models can be trained using data from these sensors to help predict the optimal beam pair. This paper proposes a novel Vision Transformer (ViT) machine-learning model for beam selection using GPS and LIDAR data. We also introduce a GPS-based Virtual Environment Capture (GVEC) solution to overcome the noise in the LIDAR data. The proposed solution outperforms previous approaches when tested on noisy LIDAR data, achieving an accuracy of 92% while searching among the top 10 beams.