RHU Computer and Communications Engineering students Ms. Aline Hassan, Ms. Aya Sharanek, and Ms. Rayan Hamzeh developed a system to optimize 5G antennas using Machine Learning (ML). They carried out their final year project under the supervision of associate professor Dr. Dina Serhal and Professor Rached Zantout. 

This project is at the forefront of the-state-of-the-art in communication systems where Machine Learning and Artificial Intelligence are being used to replace computationally intensive and trial and error methods. It is proof of RHU’s excellence in education that nurtures students’ innovation and develops their problem-solving skills.

This work provides an innovative solution to the design issues since all existing models which use ML tend to input multiple antenna dimensions and provide predictions regarding only one antenna specification. The innovative system allows a user to start with multiple specifications. It will also generate the design parameters that the user should use to realize an antenna to meet the desired specs.

Traditionally, intensive numerical computations are needed to design and optimize antennas for a specific application. The designer would start with specifications and change the dimensions and structure of an antenna by trial and error using proprietary software that costs thousands of dollars. In contrast, the system developed at the ECE department of RHU can predict the dimensions of the antenna in less than one minute using an ML model. The ML model solves the issue of having to do intensive numerical computations using proprietary and costly software. It is also an intelligent system that learns from new data to optimize future designs. Moreover, the system uses open-source software available for free.

RHU CCE students first designed and optimized a patch antenna for a mobile communication base station antenna for their final year project, then an array of patch antennas commonly used on the target base stations. They developed their own data set and trained the system to use multitude ML algorithms. They then developed a model which was verified using simulations.