GAN-based Channel Estimation for OTFS Systems

Orthogonal time frequency and space (OTFS) modulation has emerged as a promising technique for efficient and robust communication in time-varying wireless channels. The unique properties of OTFS, such as its ability to combat intersymbol interference (ISI) and leverage multipath diversity, make it an attractive choice for next-generation wireless systems. Obtaining accurate instantaneous channel state information (CSI) in advance is challenging in real communication systems. This difficulty arises due to the presence of multiple random effects within the end-to-end channel, including channel noise and variations that are either unknown or cannot be expressed analytically. We provide a unique channel estimation methodology that, in contrast to standard methods, makes use of the strength of generative adversarial networks (GANs). Specifically, we employ a conditional GAN (CGAN) based channel estimation for the OTFS (GCEO) framework to effectively map the received signal information to channel state information (CSI). By incorporating conditional information in pilot symbols, we enhance the accuracy and robustness of channel estimation in OTFS systems. The proposed GCEO framework enables the model to learn the complex mapping between the received OTFS signals and the underlying channel characteristics, thereby improving the estimation accuracy even in challenging channel conditions.