A Machine Learning Based Traveling Wave Antenna Development Methodology

A technique is developed to enable automated machine learning design of traveling wave antenna. A planar substrate is built with a reflecting ground plane and upon this an antenna strip with a number of joints is created. A machine learning algorithm is then able to vary the joint angles to reach specific goals such as the resultant reflection coefficient and gain. The outcome from this technique is a planar curl antenna with 65% bandwidth covering 3 GHz–6 GHz. The planar curl antenna is also able to provides a nearly axial beam with a target gain of 6 dBi from the zenith. This machine learning methodology can be applied to automatically develop wideband curl and spiral antenna for a large variety of applications.