Multiscale Adaptive Channel Estimation for OTFS

This paper addresses the challenge of efficient and accurate channel parameter estimation for Orthogonal Time Frequency Space (OTFS) systems in doubly selective channels. The existing approaches perform well when the channel delay and Doppler shifts are integer multiples of the delay and Doppler resolution but suffer performance degradation and require high computational complexity when dealing with a practical system with non-integer delay and Doppler shifts. We develop a novel two-stage method called Multi-scale Adaptive Channel Estimation (MACE) to overcome these limitations. Using a coarse grid search, MACE estimates the integer components of channel delay and Doppler parameters. Subsequently, a parameter reassignment rule is applied to estimate the fractional part of the delay and Doppler, enhancing accuracy. Simulation results demonstrate that MACE significantly improves the channel estimation accuracy and BER at a much lower computational complexity than existing methods.