Efficient Synthesis of Large-Scale Time-Modulated Antenna Arrays Using Artificial Neural Networks and Inverse FFT

The efficiency of beampattern synthesis for large-scale time-modulated arrays (TMAs) heavily relies on the performance of various optimization algorithms. In this article, an artificial neural network (ANN) is combined with the inverse fast Fourier transform (IFFT) to realize efficient beampattern synthesis for large-scale TMAs. Two ANNs are developed to optimize the time duration and ON–OFF time instance in each time modulation period. Specifically, the backpropagation in the ANN consists of an IFFT module and a loss function calculation module. The loss function enables to evaluate the performance of a neural network, and the ANN is optimized by minimizing the value of the loss function. The IFFT module is developed for the fast analysis of the radiation pattern of the antenna array. The desired radiation patterns could be achieved by minimizing the loss function for the radiation patterns. With this approach, radiation intensity in the main beam region could be enhanced, and the discrete form of the ON–OFF time instances and duration time in each period could be obtained as the output of the synthesis approach. A
$45\times45$
TMA is considered a numerical example to show the efficiency, flexibility, and robustness of the proposed approach.