Estimação e equalização de canal conjunta em um sistema de multiplexação espacial MIMO-OFDM por redes neurais artificiais

Abstract

Modern wireless communication faces significant challenges in dealing with complex ra dio channels, especially in urban scenarios with many obstacles and multiple paths that result in overlapping symbols, destructive interference, and echoes of transmitted signals. This leads to inter-symbol interference problems, which in turn cause bit and symbol er rors, degrading system performance. Channel estimation and equalization are the parts of signal reception responsible for understanding how the channel distorted the signal and removing this effect. The current fifth-generation mobile internet (5G) systems are powerful and complex, and their growth has allowed for advances in research and the emergence of MIMO-OFDM (Multiple Input Multiple Output Orthogonal Frequency Division Multiple xing) systems. These systems combine multiple transmitting and receiving antennas with transmission on orthogonal subcarriers to increase capacity, spectral efficiency, speed, and reliability of transmissions. In this context, the proposal of this work is to address the issue of channel estimation and equalization using artificial neural networks capable of performing these two tasks jointly in a MIMO-OFDM spatial multiplexing system, whose transmission is carried out in a channel based on a model of 5G standards. These neural networks are trai ned offline and do not use pilots. The results of this proposal show that neural networks can outperform classical techniques, including the less practical ones such as Minimum Mean Squared Error (MMSE). The complex artificial neural network (RNC) had the best Bit Error Rate (BER) results at Eb/N0 = 0 and 5 dB, while the RNAS stood out at 10 dB, maintaining a small difference compared to MMSE across the entire Eb/N0 range. The real artificial neural network (RNAR), finally, outperformed the other techniques from Eb/N0 = 15 dB. Neural networks and MMSE were analyzed in some adverse scenarios to verify the robust ness of these solutions. To simulate a scenario where the receiver terminals are moving, the Doppler effect was added, and to represent a more difficult transmission, a channel with more delays was used. To analyze the degradation caused by ISI, the transmission was carried out without the OFDM cyclic prefix. The results show that, in these adverse scenarios, the networks generally perform better than MMSE, except in the case of the Doppler effect.