Low Complexity Super-Resolution OTFS-Assisted ISAC Framework for THz Communication

Integrated sensing and communication (ISAC) in terahertz (THz) is a promising technology that supports simultaneous blue terabit-per-second data transmission and millimeter-level precision sensing. However, THz ISAC systems face significant challenges, including severe Doppler shifts and reduced power amplifier efficiency due to a high peak-to-average power ratio (PAPR). This paper presents a super-resolution orthogonal time frequency space-assisted ISAC (SR-OTFS-ISAC) framework aimed at enhancing robustness against Doppler effects in multipath THz channels. The framework effectively addresses both integer and fractional delay and Doppler. The proposed framework incorporates a low-complexity super-resolution sensing technique that begins with denoising the received signals using a super-resolution deep neural network (SR-DNN) for precise path detection. This is followed by a coarse estimation of the dominant paths using the max-path phase. This estimate is then refined using an alternative optimization (AO) method to improve the accuracy of the delay-Doppler (DD) parameters. This approach, called the super-resolution max-path alternative optimization (SR-M-PAO) method, provides better channel estimates. These enhanced estimates support multi-target sensing, transmitter localization, and data detection using a conjugate gradient algorithm. The proposed SR-OTFS design enables channel estimation, parameters sensing and data detection within a single OTFS frame. Additionally, computational complexity is derived in terms of the required real addition and multiplication operations, offering clear insights into the proposed algorithm's efficiency. Simulation results demonstrate that the SR-OTFS-ISAC framework delivers range estimation accuracy at the millimeter scale and velocity estimation precision at the centimeter-per-second level, all while ensuring lower computational complexity. Furthermore, it achieves an approximate 5 dB reduction in PAPR compared to the literature while maintaining a robust performance bit error rate, even under fractional DD effects.