Off-Grid Delay-Doppler Domain Channel Tracking for OTFS

This paper proposes a differential off-grid sparse Bayesian learning (SBL) scheme for delay-Doppler (DD) domain channel tracking in orthogonal time-frequency space (OTFS) systems. Although the complete DD domain channel becomes non-sparse and not ideally static in a high-mobility multipath channel, the differential channel between adjacent OTFS frames may exhibit sparsity when only a limited number of scatterers are moving in the propagation environment. Exploiting this differential sparsity, we formulate the DD domain channel tracking as a differential off-grid sparse signal recovery (SSR) problem. Unlike traditional channel estimation and tracking schemes, the proposed scheme exploits both temporal correlation and differential sparsity of the DD domain channel across consecutive OTFS frames, which improves the channel acquisition accuracy. An off-grid SBL channel tracking algorithm is developed, where the differential on-grid channel and the off-grid delay and Doppler parameters are iteratively updated within the SBL framework, based on the new measurement and the previously estimated channel. Numerical results demonstrate improved channel acquisition accuracy over conventional off-grid SBL channel estimation approaches and dynamic channel tracking baselines.