Low-Complexity Grant-Free Detection with Enhanced Message-Passing in LEO Satellite-IoT

Low earth orbit (LEO) satellites are expected to play an important role in enhancing terrestrial Internet of Things. This paper focuses on the challenge of joint terminal activity detection (TAD) and channel estimation (CE) in grant-free non-orthogonal random access (GF-NORA)-enabled LEO communication networks. To leverage the dominant line-of-sight (LoS) path and the limited angular spread characteristic of terrestrial-satellite links, we propose a two-stage joint TAD and CE (TS-JDE) detection scheme aimed to achieve lower computational overhead. This scheme utilizes a high-accuracy algorithm for LoS path estimation and leverages the channel correlations for low-complexity NLoS paths estimation. Specifically, in LoS estimation stage, to mitigate the inaccuracies caused by Taylor expansion errors, we propose an enhanced message-passing algorithm, termed TaMP-LoS. This algorithm exploits the structured sparsity in the delay-Doppler domain and incorporates the Lagrange remainder to evaluate the expansion error, thereby enhancing estimation accuracy. In the NLoS estimation stage, we develop a sparse Bayesian learning-based algorithm, named SBL-NLoS, designed to estimate NLoS channels characterized by the limited residuals of delay and Doppler. Simulation results show that our proposed algorithm can achieve better detection performance than the existing approaches for GF-NORA-enabled LEO networks.