PaperlayerView exact revision on arXiv

Source-linked AI summary

Fluid Antenna Systems

Kai-Kit Wong, Arman Shojaeifard, Kin-Fai Tong, Yangyang Zhang

arXiv:2005.11561v1cs.ITeess.SP

TL;DR

Fixed antenna locations can create highly correlated spaces, motivating a fluid antenna system that switches among positions to select the strongest signal. The paper derives outage-probability expressions and shows that sufficiently large N can let FAS outperform MRC even over arbitrarily small space.

  • Problem

    Fixed antenna placement can produce highly correlated spaces, motivating analysis of FAS performance and its outage-probability benefits.

  • Method

    The FAS switches the antenna to the port with the strongest signal and is analyzed through exact, approximate, and upper-bound outage-probability expressions.

  • Results

    As N →∞, FAS can achieve arbitrarily small outage probability for any positive space dimension and can outperform an L-antenna MRC system with sufficiently large N.

  • Takeaways & Limitations

    A sufficiently large number of switchable positions allows a single-antenna FAS with small space to surpass MRC under the analyzed conditions.

Abstract

from arXiv · show

Over the past decades, multiple antenna technologies have appeared in many different forms, most notably as multiple-input multiple-output (MIMO), to transform wireless communications for extraordinary diversity and multiplexing gains. The variety of technologies has been based on placing a number of antennas at fixed locations which dictates the fundamental limit on the achievable performance. By contrast, this paper envisages the scenario where the physical position of an antenna can be switched freely to one of the N positions over a fixed-length line space to pick up the strongest signal in the manner of traditional selection combining. We refer to this system as a fluid antenna system (FAS) for tremendous flexibility in its possible shape and position. The aim of this paper is to study the achievable performance of a single-antenna FAS system with a fixed length and N in arbitrarily correlated Rayleigh fading channels. Our contributions include exact and approximate closed-form expressions for the outage probability of FAS. We also derive an upper bound for the outage probability, from which it is shown that a single-antenna FAS given any arbitrarily small space can outperform an L-antenna maximum ratio combining (MRC) system if N is large enough. Our analysis also reveals the minimum required size of the FAS, and how large N is considered enough for the FAS to surpass MRC.

I. Introduction

The paper proposes a single-antenna fluid antenna system that switches among closely spaced ports to exploit spatial signal variations. It develops outage analysis and compares the system with conventional MRC under correlated fading.

  • Motivation: Closely spaced ports can still provide performance gains because small displacements may move the antenna from a deep fade to a high-reception plateau.The paper notes that displacements as small as λ/10 can make this difference.
  • System concept: Fluid antenna systems let one antenna switch among N fixed ports within a small linear space to select the strongest signal.The ports are evenly distributed, and all share one RF chain.
  • Model and objective: The model targets arbitrarily correlated Rayleigh fading and treats mutual coupling as absent because only one antenna exists.The analysis focuses on the effects of FAS space size and port count.
  • Analytical contributions: The paper derives the joint envelope distributions and an exact outage probability in single-integral form for the FAS.These expressions connect outage performance with the system’s physical parameters.
  • Analytical contributions: The approximate outage expression is closed form and is proven tight for stringent targets and strong fading correlations.An outage-probability upper bound supports further comparisons.
  • Main findings: As N approaches infinity, any positive FAS dimension can achieve arbitrarily small outage probability, while finite-N analysis quantifies when FAS surpasses L-antenna MRC.The paper also derives the minimum required FAS size for surpassing MRC.

II. System Model

The system model places one switchable antenna at N evenly spaced ports along a line of length Wλ and selects the port with the largest channel amplitude. It models the port channels as correlated Rayleigh fading with a shared single-RF-chain architecture.

  • Physical model: The FAS antenna switches among N preset locations evenly distributed over a linear dimension of length Wλ.Each port is treated as an ideal point antenna measured relative to a reference location.
  • Channel model: Each port’s channel coefficient is modeled as zero-mean circularly symmetric complex Gaussian, making its amplitude Rayleigh distributed.The model includes AWGN and a common average received SNR at the ports.
  • Correlation model: The port channels are correlated because the ports may be arbitrarily close, with spatial separation producing phase differences among arriving paths.The spatial correlation analogy follows the temporal correlation model under isotropic scattering.
  • Statistical analysis: The analysis parameterizes the correlated channels through freely chosen correlation parameters and derives distributions for the selected FAS envelope.The resulting expressions are intended to provide insight into how physical FAS parameters affect outage.
  • Selection rule: The FAS is assumed to switch instantly to the port with the maximum channel amplitude for best performance.The resulting maximum governs the analyzed FAS received signal.

III. Main Results

The section derives joint distributions and outage-probability expressions for fluid antenna systems, then develops approximations and bounds to characterize how ports, correlation, and thresholds affect performance.

  • Theorem 1 gives the joint probability density function of the magnitudes |g1|, |g2|, ..., |gN|.
  • Theorem 3 derives the FAS outage probability from the joint cumulative distribution function and the outage event.
  • Theorem 4 provides a closed-form approximation that is easier to compute and quantifies the outage reduction from additional antenna ports.
  • The approximation is tight when spatial correlation is strong or the SNR threshold relative to average SNR is stringent.
  • An additional port provides no outage reduction when it is identical to the first port, while its benefit otherwise depends on autocorrelation.

B. Physical Insights

The physical-insights analysis compares a spatially correlated-port FAS with independent-fading MRC and derives conditions linking outage performance to port count, correlation, and FAS dimension.

  • The upper bound is valid only when all factors inside its product are positive, and simulations use the bound to produce numerical results.
  • The benchmark compares a single-antenna FAS with N spatially correlated ports against an L-antenna MRC receiver with independent fading.
  • Theorem 8 states that any fixed FAS dimension can achieve arbitrarily small outage probability as N →∞ when |µk| ≠ 1.
  • A single-antenna FAS can therefore beat an L-antenna MRC system with independent fading and L RF chains in the asymptotic large-N regime.
  • For homogeneous correlation, increasing N allows µ to approach one or the required FAS space to become smaller while still outperforming MRC.
  • The general finite-N condition links the FAS dimension Wλ with system parameters and N, but valid solutions may require N to be sufficiently large.

IV. Numerical Results

Numerical results show that FAS outage performance improves with more ports and larger space, can outperform multi-antenna MRC in small spaces, and exhibits practical modeling limitations.

  • FAS scaling: Increasing N continuously reduces outage probability without an outage floor, while increasing W also reduces outage probability.The reported trend agrees with Theorem 8.
  • FAS scaling: At W = 2λ with 20 ports, outage probability reaches 2 × 10^-4, more than 4 orders of magnitude below the single-antenna baseline.
  • Approximation and bound: The approximation is accurate only at high outage probability and becomes negative as N increases, whereas the upper bound follows the decline but is loose for small W.The upper bound remains useful for linking system parameters.
  • Comparison with MRC: Even at W = 0.2, FAS outperforms 2-antenna MRC for N ≥ 7 and surpasses 5-antenna MRC when N > 70.
  • Comparison with MRC: For W = 0.2, N approaching 200 reaches outage probability 1 × 10^-5, matching 8-antenna MRC; with larger W, matching occurs at N = 23.
  • Design tradeoffs: The required N and W trade off, with 1λ FAS requiring N = 28, 61, and 102 ports to outperform 2-, 3-, and 4-antenna MRC, respectively.
  • Practical considerations: Practical interpretation is constrained because ports occupy physical space, placement resolution limits feasible N, and extreme-N performance may reflect model numerical advantages.The authors still report strong performance under practical N and W values.

VI. Conclusion

The conclusion presents FAS as a single fluid antenna switched among N ports within a linear space, with outage analysis showing strong performance even in small spaces. It also identifies practical constraints associated with the idealized port model and finite placement resolution.

  • FAS switches a single fluid antenna to the strongest of N fixed ports within a linear space of Wλ.
  • The paper derives exact and approximate outage expressions and an outage upper bound for FAS.
  • As N approaches infinity, FAS can achieve arbitrarily small outage probability for any W > 0.
  • The upper bound is used to quantify how many ports are sufficient for FAS to outperform an L-antenna MRC system.
  • The conclusion states that small-space FAS with practically feasible N can outperform MRC, while practical aspects require further investigation.

A. Derivation of p|g1|,|g2|,...,|gN|(r1, r2, . . . , rN)

The derivation conditions on one channel magnitude, uses conditional Rician distributions and conditional independence, and integrates the resulting joint density to obtain the target result.

  • Conditioned on x0 and y0, |g2| is Rician distributed.
  • Given x0, y0, the magnitudes |g2| through |gN| are conditionally independent.
  • The desired joint-density result is obtained by integrating the conditional density of |g2| through |gN| with the density of |g1|.
  • The joint pdf of |g1|, |g2|, ..., |gN| is formed through nested integration and substitution into the preceding expression.
  • The inner product integral is recognized as integration over a Rician pdf, and a variable change yields the final result (13).

C. Derivation of the Joint cdf when N = 2

This section develops the N = 2 joint-cdf analysis underlying the FAS outage-probability approximation. It uses bounds and monotonicity properties of Q1-related terms to assess when the approximation is tight.

  • Joint-cdf derivation: The derivation uses Lemma 3 and a change of variables to obtain the needed integral result.The result is substituted into the preceding expression to complete the derivation.
  • Joint-cdf derivation: Q1(α, β) is upper-bounded for 0 ≤ α ≤ β, with erfc(·) and I0(x) bounds applied to obtain the desired result.The bounds use the complementary error function and a modified-Bessel-function inequality.
  • Tightness analysis: The auxiliary function f(x) decreases initially and may later increase, so its maximum occurs at an endpoint of [0, X0].The endpoint analysis uses the derivative behavior of f(x), together with f(0) = 1 and f(X0) = 1.
  • Tightness analysis: The approximation is based on Q1(αk, βk) terms and is accurate when these terms are small.The analysis identifies small Q1(αk, βk) as the condition supporting approximation accuracy.
  • Tightness analysis: The upper bound remains near one only over an insignificant integration interval and then falls sharply under the stated conditions.Consequently, the upper bound is small over a significant part of the integration range, making the approximation tight.
  • Tightness analysis: The resulting approximation is therefore tight when the upper bound on Q1(αk, βk) is small over a significant integration interval.This conclusion completes the tightness argument for the outage-probability approximation.

F. Derivation of the Upper Bound, pUB

This section derives an outage-probability upper bound by lower-bounding the change in outage probability. The derivation applies a lower bound on Q1(α, β) for large β.

  • Upper-bound derivation: The upper bound pUB_out(γth) is obtained by first deriving a lower bound for Δpout(γth).The outage probability for an n-port FAS is represented using the superscript n.
  • Upper-bound derivation: Lemma 6 supplies a lower bound for Q1(α, β) when 0 < α < β and β is large.The bound introduces κ as any positive constant greater than one and defines ϱ accordingly.
  • Upper-bound derivation: Applying the lower bound from Lemma 6 and a Gaussian Q-function bound yields the stated outage-probability upper bound.The proof concludes after substituting these bounds into the preceding expression.

G. The Minimum Dimension of Fluid Antenna, W

This section uses the outage upper bound to derive the minimum FAS dimension needed to obtain a target autocorrelation and outperform MRC. It then extends the result to general port spacings through a sufficient condition.

  • Minimum dimension: Corollary 7 derives the minimum dimension d* needed to achieve the required autocorrelation μ* for FAS to outperform MRC.The result is based on the upper bound of outage probability for a system whose autocorrelation parameters are equal.
  • Minimum dimension: If half the ports have dk < d* and the other half have dk > d*, the FAS has better outage-probability performance than the corresponding equal-autocorrelation case.The equal-autocorrelation N/2-port FAS is used as a worst-case reference for the general N-port system.
  • Minimum dimension: The equal-autocorrelation result provides a sufficient condition for the required FAS dimension Wλ in the general-{μk} setting.The condition follows from the result of Corollary 7.
Loading 2005.11561v1…