Machine Learning Based Fully Digital UWB Antenna for Direction Finding Systems

This work presents a new generation of ultra-wideband (UWB) radio frequency direction-finding system. The architecture of such a system is based on phase interferometry and exploits all leading-edge technology solutions such as the direct sampling of the entire analog band or the introduction of Artificial Intelligence for the processing of incoming RF signals. Thanks to these new solutions, a minimum number of antennas is needed to cover a multi-octave band capable to operate in the so-called folded mode. The presented solution is based on four full-band EM interferometer antenna array. The same signal collected from each antenna, after a first analog treatment, is then digitized with different sampling frequencies to get the diversity required for solving the frequency ambiguity problem. The Machine Learning approach is then adopted to realize the direction of arrival estimation. Comparisons between standard processing techniques and ML approach confirm the effectiveness of the presented solution.