Off-Grid Channel Estimation Using Grid Evolution in Rectangular Pulse-Shaping OTFS System

Orthogonal time frequency space (OTFS) can prvide better error performance than orthogonal frequency division modulation in the high-speed scenario. Recent studies focus on off-grid channel estimation under the ideal shaping pulse. However, the utilization of the practical rectangular shaping pulse degrades the performance of the approaches, which rely on the ideal pulse to decouple the delay and the Doppler. In this paper, an off-grid channel estimation algorithm is proposed based on the introduction of the grid evolution in the sparse Bayesian inference named the GESBI algorithm. Specifically, the grid evolution is introduced in the external iteration to update the virtual grid through the estimated on-grid and off-grid elements in the internal iteration. Different from the fixed and uniform virtual grid in the previous off-grid channel estimation algorithms, the virtual grid in the GESBI algorithm is non-uniform with the grid evolution. In addition, the delay-Doppler domain input-output relation is derived indicating the invalidity of the existing low-complexity channel estimation algorithms under the rectangular pulses. In the GESBI algorithm, a special design is made for the rectangular pulse to improve the off-grid channel estimation accuracy. Finally, simulation results demonstrate the efficiency of the proposed GESBI algorithm.