Antenna beamforming for EW using adaptive layered networks

Artificial neural networks (ANNs) provide an alternative `inverse processing´ solution for angle-of-arrival (AOA) estimation in the form of an adaptive layered network. A suitably trained supervised ANN, such as the multi-layer perceptron, can classify signal AOA into a number of sectors with processing input taken directly from the receiving array elements. The main problems are that uniform angular sectoring is not compatible with linear array functions, and unless the output space is carefully code-mapped, rays with AOA near the boundaries will give large estimation errors. These and other problems were resolved by modifying the ANN method using least squares and a novel graphical solution together with unit distance (Gray) coding