Beyond Traditional Techniques: Utilizing Neural Networks to Improve and Compare the Performance of OFDM and OTFS

Abstract Two popular modulation techniques for current 4G, 5G, and beyond wireless communication systems are Orthogonal Frequency-Division Multiplexing (OFDM) and Orthogonal Time Frequency Space (OTFS). While OFDM performs quite well in static channels, it suffers greatly in situations with excessive mobility. On the other hand, OTFS operates in the delay-Doppler domain providing robustness in these kinds of high-mobility channels. The potential of neural network (NN) models for OFDM and OTFS performance improvement and comparison analysis is investigated in this work. We proposed a framework for analyzing and contrasting the effectiveness of both modulations in different NN models. To compare and enhance the performance of OFDM and OTFS systems, this paper explores the wide range of Neural Network (NN) models, such as feedforward Neural Networks (NNs), Recurrent Neural Networks (RNNs) with particular architectures like Gated Recurrent Units (GRUs), and Long Short Term Memory (LSTM) networks. Out of the four proposed models, the RNN model promises better performance when compared to the rest and the traditional method which is the non-NN model. Furthermore, the proposed framework optimizes some features of OFDM and OTFS by utilizing NNs’ learning capabilities, which results in higher performance. The results show that this strategy can improve wireless communication performance, especially in dynamic channel conditions.