Agatonovic M.   Stankovic Z.   Milovanovic B.   Doncov N.  

An approach in linear antenna array synthesis based on artificial neural network modeling

Reporter: Agatonovic M.

The capacity of wireless communication systems can be substantially increased by the use of space division multiple access (SDMA) technique. This channel access method exploits spatial separation between different users in order to suppress unwanted interference signals. Antenna arrays have crucial role in the realization of such space filtering. For that purpose, effects of antenna array elements spatial distribution on the radiation pattern are investigated in this paper. Linear antenna array consisting of four by one (4x1) microstrip patch antennas at operating frequency of 2.4 GHz is considered. If the synthesis process is performed using one of the available EM solvers, it can be computationally expensive and time-consuming. A demand for new design methodologies that are both accurate and fast at the same time has appeared. An alternative solution to overcome the existing problem is to employ an artificial neural network (ANN) model, where an EM solver is utilized only to generate training and test data for the model. Employing both EM solver and the neural model, optimal inter-element spacings between array elements that provide minimum values of sidelobe level (SLL) and desired half-power beamwidth (HPBW), are determined. In comparison with the duration of classical synthesis process using only the EM solvers, the total time required to obtain data for developing ANN model and to perform synthesis process using this ANN model is significantly reduced.