Using radial-basis function neural networks to shape the array factor and reduce the side-lobe levels of phased antenna arrays

Abstract

Shaping the array factor of an adaptive antenna array to obtain interference suppression is a difficult task due to the computational complexity, slow convergence rates and the high cost requirements. Although some antenna synthesis techniques can be used to reduce the side-lobe levels (and hence reduce the effect of interference arriving outside the main lobe), the resulting array factors suffer from the increased width of the main lobe. This degradation is more profound in phased arrays operating at large scanning angles and hence, the performance will be limited in many applications requiring radiation patterns with narrow steerable main lobes. In this paper, a technique based on radial-basis function neural networks (RBFNN) is presented for shaping the array factor of phased linear arrays to have relatively low side-lobe levels without affecting the beamwidth requirements of the main lobe. Both uniform and nonuniform linear arrays with initially low sidelobe levels, such as Tschebyscheff arrays are considered. The simulation results show that the use of RBFNN minimizes the sidelobe levels while keeping a predetermined width of main lobes. Thus, highly improved patterns with very deep sidelobe and increased directivity, with beam steering capabilities, are obtained.