Complex MIMO RBF Neural Networks for Transmitter Beamforming over Nonlinear Channels

Area of Science:

• Electrical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Beamforming is crucial for enhancing spectral and energy efficiencies in next-generation wireless systems and low Earth orbit satellites.
• Practical systems already utilize beamforming, which significantly increases capacity and minimizes spectral congestion.

Purpose of the Study:

• To propose a novel complex multiple-input multiple-output radial basis function neural network (CMM-RBF) for transmitter beamforming.
• To evaluate the performance of the proposed CMM-RBF against the least mean square (LMS) algorithm.

Main Methods:

• Developed a CMM-RBF based on the phase transmittance radial basis function neural network (PTRBFNN).
• Compared CMM-RBF with the LMS algorithm using antenna arrays (six dipoles in a uniform circular array and 16 dipoles in a 2D-grid array).

Main Results:

• The proposed CMM-RBF demonstrated a lower steady-state mean squared error compared to the LMS algorithm.
• CMM-RBF exhibited a faster convergence rate than the LMS algorithm.
• Enhanced half-power beamwidth (HPBW) was observed with the CMM-RBF in nonlinear scenarios.

Conclusions:

• The novel CMM-RBF offers superior performance for transmitter beamforming in nonlinear scenarios.
• CMM-RBF presents a promising advancement for improving efficiency in wireless communication systems.
• The proposed method achieves better error reduction, faster adaptation, and narrower beamwidth compared to traditional LMS algorithms.