Deep learning for fractional channel estimation in OTFS receivers

The rapid advancement of mobile communication technologies has necessitated innovative solutions for channel estimation, particularly in high-mobility scenarios where traditional methods struggle to maintain performance. This thesis addresses the challenges of fractional channel estimation in Orthogonal Time Frequency Space (OTFS) systems, a candidate waveform for 6G networks. This work begins with a comprehensive review of wireless communication channels and existing channel estimation techniques for OTFS systems, highlighting their limitations in dealing with inter-path interference (IPI) in environments with fractional delays and Doppler shifts. To address these challenges, a variant of a state-of-the-art algorithm, named Progressive Inter-path Interference Cancellation (P-IPIC), is proposed to significantly reduce computational complexity and latency while improving estimation accuracy. Specifically, the results demonstrate that a single refinement procedure is sufficient to achieve good estimation performance. Additionally, the integration of a deep learning (DL) approach into the channel estimation process is explored. A neural network model is developed to learn the intricate relationship between channel parameters and a corresponding parameter matrix. Simulation results show that while deep learning solutions contribute to latency reduction, they come with a minor, yet acceptable, performance loss. The findings of this thesis offer valuable insights and practical solutions for next-generation wireless communication systems, particularly in the context of 6G technology.