Neural Modeling For Synthesis of Sector Beam and Multibeam Array Antennas

A new and very speedy low-sidelobe pattern synthesis method for linear array antennas with periodic element spacing is described. This paper presents an efficient method for the pattern synthesis based on neural network modelling and Fourier transform method. The proposed synthesis method provides important improvements in terms of performance, computational speed, elasticity, and ease of implementation in software to the methods described in literature. A number of representative examples are presented to demonstrate the various unique capabilities of the method. The results include sum and difference patterns for antenna array. The approach combines a neural network with the hardware limitations of the array to place nulls in the directions of interference with small perturbations to the far-field pattern. Excellent nulling results are possible for most interference scenarios.