[Paper Review] User Activity Detection and Channel Estimation for Grant-Free Random Access in LEO Satellite-Enabled Internet-of-Things

Zhaoji Zhang, Ying Li+5 more|arXiv (Cornell University)|May 20, 2020

IoT Networks and Protocols36 references4 citations

TL;DR

This paper proposes a Bernoulli-Rician message passing with expectation maximization (BR-MP-EM) algorithm for joint user activity detection (UAD) and channel estimation (CE) in grant-free random access (GF-RA) for low-Earth orbit (LEO) satellite-enabled Internet-of-Things (IoT) networks. The method enables low-latency, robust access by iteratively refining activity and channel estimates through message passing and EM-based hyperparameter updates, achieving high accuracy even under severe channel impairments.

ABSTRACT

With recent advances on the dense low-earth orbit (LEO) constellation, LEO satellite network has become one promising solution to providing global coverage for Internet-of-Things (IoT) services. Confronted with the sporadic transmission from randomly activated IoT devices, we consider the random access (RA) mechanism, and propose a grant-free RA (GF-RA) scheme to reduce the access delay to the mobile LEO satellites. A Bernoulli-Rician message passing with expectation maximization (BR-MP-EM) algorithm is proposed for this terrestrial-satellite GF-RA system to address the user activity detection (UAD) and channel estimation (CE) problem. This BR-MP-EM algorithm is divided into two stages. In the inner iterations, the Bernoulli messages and Rician messages are updated for the joint UAD and CE problem. Based on the output of the inner iterations, the expectation maximization (EM) method is employed in the outer iterations to update the hyper-parameters related to the channel impairments. Finally, simulation results show the UAD and CE accuracy of the proposed BR-MP-EM algorithm, as well as the robustness against the channel impairments.

Motivation & Objective

• To address the challenge of sporadic, random access from massive IoT devices in LEO satellite networks with high latency and limited feedback.
• To enable grant-free random access (GF-RA) in LEO satellite-IoT systems to reduce access delay and signaling overhead.
• To jointly detect active users and estimate their channels in the presence of Rayleigh and Rician fading, plus phase and amplitude impairments.
• To develop a robust, iterative algorithm that adapts to unknown channel impairments without prior knowledge.

Proposed method

• Proposes a two-stage BR-MP-EM algorithm: inner iterations perform Bernoulli-Rician message passing for joint UAD and CE, while outer iterations use expectation maximization (EM) to refine hyperparameters.
• Uses message passing on a factor graph to compute posterior estimates of user activity and channel state, incorporating Rician fading and Rayleigh scattering components.
• Employs EM updates in outer iterations to estimate unknown channel impairment parameters, such as amplitude gain $ h_r $ and phase shift $ heta_\delta $, based on iterative likelihood maximization.
• Modifies the message update rules for the case where impairments affect only the LoS component, replacing time-varying variances with constant $ v^{\text{ray}}_k $.
• Derives closed-form solutions for $ \hat{h}_r $ and $ \hat{\phi}_\delta $ using weighted averages of complex channel estimates, ensuring physical consistency by selecting the correct solution branch.
• Validates the algorithm using simulation results under realistic LEO satellite-IoT channel models with mixed LoS and non-LoS components.

Experimental results

Research questions

• RQ1 How can user activity detection and channel estimation be jointly optimized in a grant-free random access scheme for LEO satellite-enabled IoT?
• RQ2 What is the performance limit of UAD and CE when channel impairments (amplitude and phase) are unknown and affect both LoS and NLoS components?
• RQ3 Can message passing with EM-based hyperparameter learning achieve robust and accurate estimation under realistic LEO satellite channel conditions?
• RQ4 How does the proposed BR-MP-EM algorithm compare to conventional methods in terms of UAD and CE accuracy under varying SNR and channel impairment levels?

Key findings

• The proposed BR-MP-EM algorithm achieves high user activity detection accuracy, correctly identifying active users even in low SNR regimes with high interference.
• Channel estimation accuracy is significantly improved, with the algorithm converging to accurate estimates of both $ h_r $ and $ \phi_\delta $, the key parameters of channel impairments.
• The algorithm demonstrates robustness against unknown channel impairments, maintaining high performance even when the true values of $ h_r $ and $ \phi_\delta $ are not known a priori.
• Simulation results show that the BR-MP-EM algorithm outperforms conventional methods in both UAD and CE, particularly in scenarios with mixed LoS and NLoS propagation.
• The modified version of the algorithm, when impairments affect only the LoS component, achieves a closed-form solution for $ \hat{h}_r $ and $ \hat{\phi}_\delta $, improving computational efficiency.
• The physical interpretation of the solution ensures that only the correct branch of the two possible solutions is selected, based on the expected phase alignment of channel estimates.