Machine Learning Enhancement of a Luneburg Lens Amplitude-Only Direction-Finding Antenna

A 5-element cylindrical Luneburg lens antenna, operating at 27.5 GHz, is proposed for a receive-only application. The array is poised to be integrated with amplitude-only receivers, for the purpose of determining the angle-of-arrival of a single signal received by the array. A conventional methodology for performing this task is shown as a baseline with machine-learning solution introduced for comparison. The peak improvement of the machine-learning technique shows an accuracy increase of up to 41%. In situations with reduced signal-to-noise ratio, the performance of the neural network solution outperforms the conventional method, demonstrating an improvement of 30%.