Joint Timing Offset Estimation, User Activity Detection, and Channel Estimation for Asynchronous OTFS-Based Grant-Free Random Access

Abstract

Grant-free random access (GFRA) is a key enabler for massive machine-type communications (mMTC) in future wireless networks. However, supporting massive access under high mobility and asynchronous transmissions remains a fundamental challenge, especially for orthogonal time–frequency space (OTFS) modulation, where timing offsets severely disrupt the structured sparsity of the delay–Doppler channel representation. This work proposes an efficient asynchronous OTFS-based GFRA framework that enables reliable massive access without strict time synchronization. By leveraging a basis expansion model (BEM), we characterize the time-varying channel in a low-dimensional form, and formulate the joint timing offset estimation, user activity detection, and channel estimation as a structured compressive sensing problem. A bi-level sparsity structure is identified and exploited, consisting of common sparsity across multiple receive antennas and delay-constrained sparsity across mMTC users. To effectively exploit this structure, we develop a multi-layer factor graph and propose a novel multi-layer expectation propagation (MLEP) algorithm to enhance joint estimation performance. We also develop a computationally efficient multi-layer approximate message passing (MLAMP) algorithm to achieve near-optimal performance with reduced complexity. Extensive simulation results demonstrate that the proposed schemes significantly outperform existing benchmark schemes, with the MLEP algorithm approaching the Oracle minimum mean square error (MMSE) performance bound.