وسام الخالدي

In modern wireless communications, the robustness of the signal against dynamic an‎d noisy channel conditions is of ever-increasing importance. In this study, we describe an equalization solution based on an Artificial Neural Network (ANN1) structured around a MIMO2-OFDM3 delivery system, whereby the focus is on the accuracy of detection, an‎d also on robustness against distortion of the channel. Here we have developed an intelligent receiver designed to obtain a complete learning of the nonlinear signal transformation rather than using the stan‎dard analytical methods of continued inversion of methodology for equalizers. The proposed structure has been entirely implemented in the Python programming environment using the PyTorch frameworks, which allowed itself integration of practical data of channel effect behavior, in particular, Rayleigh an‎d Rician fading, Additive White Gaussian Noise (AWGN4), Doppler shift, an‎d Carrier frequency offset (CFO5) bit reference feedback data to be included. Hence, realistic behavior of the system may be expected. The ANN has been trained over 180 epochs, representative of the 256 hidden neurons of applied ReLU activation an‎d dro‎pout methodology of regularization. Simulations, through a wide variation of Eb/N0 conditions, were then used to indicate capabilities for generalization an‎d adaptability. The results of experimental simulation indicate very successful advances have been made whereby the Bit Error Rate (BER6) performance of the ANN equalizing systems are very good with respect to overall BER performance in conditions whereby when compared with stan‎dard ZF7, MMSE8 an‎d MMSE-SIC9 methods 1.5-2 dB of improvement is attained with also 96.8% success in detection over a large number of channel configurations. Such results indicate very apparent that such an ANN equalization methodology based on learnt data thereby improves accomplished robustness of detection errors, an‎d at the same time gives methods of operational methodology an‎d inherent computational efficiency by the techniques through the ANN neural network derived methodologies, potentially leading to their methods of implementation in real-time operational performance systems.Overall, this research will clearly strongly demonstrate the case for ANN-derived equalization systems in operation, a serious scope for further detailed research in their own right, an‎d subsequently point out to structures which are likely to be a major advance towards intelligent an‎d adaptive receiver systems in subsequent 5G10/6G11 communication systems to come.

كنترل هوشمند , شبكه‌هاي عصبي مصنوعي (ANN) , اينترنت اشيا IoT , مدولاسيون ديجيتال , چند ورودي-چند خروجي MIMO , هوش مصنوعي

Smart Control , Artificial Neural Networks ANN , Internet of Things IoT , Digital Modulation , Multi-Input-Multi-Output MIMO , Artificial Intelligence AI

WISAM ALKHALIDI

Dr.Abolfazl Diyanat