A FFNN-Based System for the DoA Estimation in 6G Massive Wireless Communications: Analysis and Evaluation

In this work, a well-developed system, consisting of a transmitter and receiver for DoA estimation, has been proposed to overcome challenges of non-linearity, complexity, and heterogeneity in 6G massive wireless communications. By increasing the matching between the transmitter and receiver of the system, we can significantly enhance its performance at an operating frequency of 100 GHz. Hence, the HAAwBE with a very directive radiation pattern has been implemented in the transmitter. In addition, a robust FFNN algorithm has been implemented in the beamforming receiver. The beamforming sys-tem is capable of fully suppressing interference, while efficiently detecting a signal of interest. Moreover, we have evaluated system performance using the regression metric, training performance, error histogram and a validation dataset. The R-values of the training, validation, testing, and all (i.e, the complete) dataset are equal to 0.9995, 0.9983, 0.9996, and 0.9995, respectively, which are accompanied by a highly reliable and accurate beamforming system. Also, the best validation performance shows a very low MSE of
$5.32\times 10^{-4}$
. Moreover, the training state diagram of the FFNN model has reported a gradient of 0.028571, a Mu value of
$1\times 10^{-5}$
, and validation checks of zero after passing 10 epochs.