PaperlayerView exact revision on arXiv

Source-linked AI summary

Channel Estimation for Wireless Communication Systems Assisted by Large Intelligent Surfaces

Junliang Lin, Gongpu Wang, Rongfei Fan, Theodoros A. Tsiftsis, Chintha Tellambura

arXiv:1911.02158v1cs.ITeess.SPmath.IT

TL;DR

LIS-assisted channel estimation must account for assistant-channel features that make the task differ from traditional estimation. The paper formulates constrained error minimization, solves it with an iterative Lagrange multiplier and dual-ascent estimator, and derives CRLBs. Simulations report up to 18% higher estimation accuracy than LS in the low-SNR regime.

  • Problem

    Assistant-channel features had not been incorporated into channel estimation for LIS-assisted wireless systems, motivating a constrained formulation.

    Source: §I · p. 1, §II · p. 2

  • Method

    The paper uses a Lagrange multiplier and dual-ascent scheme to obtain an iterative closed-form estimator and derives corresponding CRLBs.

    Source: §I · p. 1, §III · p. 2, B. Cram´er-Rao Lower Bound · p. 3

  • Results

    At SNR = 0 dB, DES improves estimates of ℜ(h), ℑ(h), and η over LS by 16.53%, 18.12%, and 8.24%, respectively.

    Source: §IV · p. 3

  • Takeaways & Limitations

    The proposed DES achieves at most 18% higher estimation accuracy than LS in the low-SNR regime.

    Source: §V · p. 4

Abstract

from arXiv · show

In this letter, the channel estimation problem is studied for wireless communication systems assisted by large intelligent surface. Due to features of assistant channel, channel estimation (CE) problem for the investigated system is shown as a constrained estimation error minimization problem, which differs from traditional CE problems. A Lagrange multiplier and dual ascent-based estimation scheme is then designed to obtain a closed-form solution for the estimator iteratively. Moreover, the Cramer-Rao lower bounds are deduced for performance evaluation. Simulation results show that the designed scheme could improve estimation accuracy up to 18%, compared with least square method in low signal-to-noise ratio regime.

I. INTRODUCTION

LIS-assisted communication can reduce energy consumption and hardware expenditure, but channel estimation must account for assistant-channel features that prior work had not investigated. This letter formulates that estimation task as constrained optimization and develops an iterative estimator with CRLB-based evaluation.

  • LIS uses configurable low-cost reflecting elements to adjust incident-signal phases with low energy consumption using channel state information.Source: §I · p. 1
  • Prior LIS channel-estimation studies used binary reflection control, time division with minimum mean squared error, or compressive sensing and deep learning.Source: §I · p. 1
  • Earlier approaches estimated channel parameters sequentially or relied on large pre-collected training datasets.Source: §I · p. 1
  • The impact of assistant-channel features on channel estimation had not been investigated, motivating this work.Source: §I · p. 1
  • The proposed formulation incorporates assistant-channel features as constrained optimization, derives an iterative closed-form estimator, and evaluates it with CRLBs and LS comparisons.Source: §I · p. 1

II. SYSTEM MODEL AND CHANNEL MODEL

The system combines direct and LIS-assisted links, with pilot observations governed by whether the LIS reflects each symbol. The estimator targets h and η while enforcing positivity and assistant-channel superiority constraints.

  • The model contains a single-antenna BS and UE, a controller, and an LIS with N reflecting elements assisting downlink transmission.Source: §II · p. 1
  • The direct BS–UE channel is h, while f and g represent the BS–LIS and LIS–UE channel vectors with complex Gaussian fading elements.Source: §II · p. 1
  • The indicator r[k] distinguishes unreflected symbols, r[k] = 0, from LIS-reflected symbols, r[k] = 1, producing different received-signal forms.Source: §II · pp. 1–2
  • The phase-shift matrix Φ configures reflecting-element phases, while the assistant channel is simplified using amplitude vectors and phase cancellation ΘΦΨ = I under perfect CSI.Source: §II · p. 2
  • The channel parameters h and η are estimated subject to η > 0 and η > |h|, representing assistant-channel positivity and superiority over the direct channel.Source: §II · p. 2

III. CHANNEL ESTIMATION

The channel-estimation section develops a Lagrange-multiplier and dual-ascent scheme for the constrained problem, then derives the corresponding CRLB for evaluation.

  • A Lagrange multiplier and dual-ascent estimation scheme addresses the constrained channel-estimation problem.The scheme is designed to obtain a closed-form estimator iteratively.Source: §III · p. 2
  • The corresponding Cramér-Rao lower bound is derived to evaluate estimator performance.Source: §III · p. 2

A. Scheme Design

The scheme separates reflected and unreflected pilots, formulates constrained estimation in a convexified signal model, and solves it through Lagrange multipliers and iterative dual ascent.

A. Scheme Design

  • Each slot contains Kp pilot symbols and Kd data symbols, with K1 unreflected and K2 reflected pilots.Source: A. Scheme Design · p. 2
  • Pilot sets K1 and K2 collect transmitted symbols, noises, and received signals for the two reflection states.Source: A. Scheme Design · p. 2
  • The pilot observations are reconstructed in vector form and represented through the pilot matrix, channel-parameter vector, noise vector, and received-signal vector.Source: A. Scheme Design · p. 2
  • The CE problem is formulated as constrained estimation-error minimization using assistant-channel features, including C2: η > |h|, which is squared to obtain C2′: xHCx ≤ 0.Source: A. Scheme Design · p. 2
  • After convexification, the method constructs a Lagrange function with nonnegative multipliers and uses KKT conditions to obtain a closed-form optimal estimator.Source: A. Scheme Design · pp. 2–3
  • Dual ascent updates the multipliers and estimator sequentially until convergence, using iteration-dependent stepsizes and projection onto nonnegative values.Source: A. Scheme Design · p. 3

B. Cram´er-Rao Lower Bound

The paper derives CRLBs for the channel parameters by expressing the Fisher information matrix under a real-valued parameterization and uncorrelated complex Gaussian noise.

IV. NUMERICAL RESULTS

Simulations compare DES with LS and the derived CRLB across SNR. DES improves estimation accuracy most at low SNR, remains above the CRLB, and converges rapidly through its multiplier updates.

  • The simulations average results over 10^4 independent Monte Carlo trials with N = 32 reflecting elements and Kp = 2 pilot symbols.Source: §IV · p. 3
  • DES achieves lower MSE than LS, and its MSE approaches the corresponding CRLB as SNR becomes sufficiently high.Source: §IV · p. 3, Figure 3 · p. 4
  • 18.12% is the largest reported DES accuracy gain over LS at SNR = 0 dB, while gains are around 2% at SNR = 8 dB.Source: §IV · p. 3, Table I · p. 4
  • DES remains above the CRLB across the evaluated SNR range, supporting the validity of the derived theoretical bounds.Source: §IV · p. 4, Figure 3 · p. 4
  • The Lagrange multipliers become constant after approximately 10 and 16 iterations, indicating rapid convergence to globally optimal estimates.Source: §IV · p. 4

V. CONCLUSION

The conclusion frames LIS-assisted channel estimation as constrained error minimization, solves it iteratively with a Lagrange multiplier and dual ascent scheme, and reports up to 18% accuracy improvement over LS at low SNR.

  • LIS-assisted channel estimation is formulated as constrained estimation-error minimization because of assistant-channel features.Source: §V · p. 4
  • The proposed scheme iteratively obtains a closed-form expression for the combined channel parameters and uses CRLBs for evaluation.Source: §V · p. 4
  • 18% is the maximum reported DES accuracy improvement over LS in the low-SNR regime.Source: §V · p. 4
Loading 1911.02158v1…