A new technique for linear antenna array processing for reduced sidelobes using neural networks

In adaptive antenna arrays (AAA), shaping the array factor is a challenging task where different sophisticated adaptation algorithms might be utilized. Applying these algorithms to AAA results in limited performance, due to slow convergence rates and increased computational complexity. The high cost required is another factor that should be taken into consideration. From another point of view, the synthesis of linear arrays to produce highly reduced sidelobes is a problem of similar complexity and limitations. This paper suggests a new synthesis technique for antenna array systems based on a trained neural network (NN). In particular, the output of a linear array is processed by two NNs, The simulation results show that the NN, when trained to minimize the sidelobe levels of the array, results in highly improved patterns with very deep sidelobes. This method substitutes other tedious conventional algorithms that are usually used in adaptive antenna arrays and array synthesis.