PaperlayerView exact revision on arXiv

Source-linked AI summary

A Geometric Interpretation of Fading in Wireless Networks: Theory and Applications

Martin Haenggi

arXiv:0711.1890v1cs.ITmath.IT

TL;DR

For block fading, the paper asks which nodes receive a packet at least once over n transmissions. It develops broadcast reliability and transport-gain analyses; under Rayleigh fading, the average connected node is 1+δ times farther away than without fading.

Abstract

from arXiv · show

In wireless networks with random node distribution, the underlying point process model and the channel fading process are usually considered separately. A unified framework is introduced that permits the geometric characterization of fading by incorporating the fading process into the point process model. Concretely, assuming nodes are distributed in a stationary Poisson point process in $\R^d$, the properties of the point processes that describe the path loss with fading are analyzed. The main applications are connectivity and broadcasting.

I. INTRODUCTION AND SYSTEM MODEL

Because distance and fading can both be uncertain, the paper proposes treating them jointly in a point-process framework. Its system model combines a stationary Poisson point process with unit-mean fading to form a path-loss process with fading.

A. Motivation

  • The framework treats distance uncertainty as large-scale fading and multipath uncertainty as small-scale fading, using Nakagami-m to include Rayleigh fading (m=1) and no fading (m→∞).Source: A. Motivation · pp. 1–2

B. Notation and symbols

  • Slanted sans-serif symbols denote random variables, and the paper uses “points” and “nodes” interchangeably.Source: B. Notation and symbols · p. 2

C. Poisson point process model

  • Nodes are modeled as a stationary unit-intensity Poisson point process, transformed by the path-loss exponent and iid unit-mean fading into a path-loss process with fading.Source: C. Poisson point process model · pp. 3–4
  • A node is connected to the origin when its faded path loss is below 1/s, and the paper defines counting measures and densities for the relevant point processes.Source: C. Poisson point process model · p. 4

D. The standard network

The section characterizes path-loss point processes under fading, including a distribution-preserving case when the path-loss exponent equals the network dimension. It also examines connectivity and how fading reorders nodes.

D. The standard network

  • The standard network uses δ=1 and Rayleigh fading; its illustration marks nodes reachable from the center against the non-fading transmission range.The example uses a path-gain threshold of s=0.1.Source: D. The standard network · p. 4, Figure 1 · p. 5

II. PROPERTIES OF THE POINT PROCESSES

  • Φ, Ξ, and ˆΞ are Poisson point processes.The framework maps geographic distances into path loss including fading.Source: §II · pp. 5, 8
  • For rational m, d, and α, the distribution Fξi can be expressed using hypergeometric functions.As m→∞, the cited distribution approaches that of x1.Source: §II · pp. 6–7
  • When δ=1, any unit-mean fading distribution preserves the distribution of the point process.Without geographic distance information, a receiver cannot determine how much fading is present.Source: §II · pp. 7–8
  • 1/e of nodes in any interval [0, a) leave it under Rayleigh fading, and the expected number entering equals the number leaving.The equal expected inflow and outflow applies to Nakagami-m fading when δ=1.Source: §II · p. 8
  • 1/(j + 1) is the limiting probability that a node has the largest fading coefficient among j + 1 nodes as its index i grows.For equal rank k, Pk,k tends to 1/3 as k grows.Source: §II · p. 10

III. CONNECTIVITY

This section investigates the processes describing path loss and connected nodes.

III. CONNECTIVITY

  • The analysis considers the processes ˆΦ and ˆΞ = Ξ ∩ [0, 1/s) for connected nodes.Nodes are considered connected when their path loss is below the threshold 1/s.Source: §III · p. 11

A. Single-transmission connectivity and fading gain

For a single transmission, connected nodes form a thinned point process, and the effect of fading on their number depends on the path-loss exponent relative to network dimension. The analysis also characterizes isolation probability and the distances of connected nodes.

A. Single-transmission connectivity and fading gain

  • With fading, connected nodes form a Poisson process with intensity λ(x)(1 − F(sx)).The number of connected nodes is Poisson; in the standard network, its mean is c_d/s and the isolation probability is e^(−c_d/s).Source: A. Single-transmission connectivity… · pp. 11–12
  • When the path-loss exponent equals network dimension (δ = 1), the expected number of connected nodes is independent of the fading type or its presence.For Nakagami-m fading, the result is independent of m.Source: A. Single-transmission connectivity… · pp. 11–12
  • The connectivity fading gain equals the δ-th moment of fading and can increase or decrease the expected number of connected nodes depending on δ.For Nakagami-m fading it decreases with m when δ > 1, increases when δ < 1, and equals 1 for all m when δ = 1; for Rayleigh fading at δ = 1/2, the gain is π/2.Source: A. Single-transmission connectivity… · p. 12
  • In the standard network, the probability of isolation is e^(−c_d/s).The number of connected nodes is Poisson with mean c_d/s.Source: A. Single-transmission connectivity… · pp. 11–12
  • For Rayleigh fading, the average connected node is 1 + δ times farther away than without fading.Without fading, the expected distance is δ/(s(δ + 1)); for Nakagami-m fading, the mean is 1 + δ/m times the no-fading value.Source: A. Single-transmission connectivity… · pp. 12–13

B. Connectivity with retransmissions

With iid block fading, the paper characterizes the nodes that receive a packet at least once over repeated transmissions. In the standard network, the expected number reached grows with the number of transmissions, and a decreasing threshold schedule yields linear growth.

B. Connectivity with retransmissions

  • With iid block fading, the process of nodes receiving at least one of n transmissions has density λ_n(x).The density is given by λ_n(x) = (1 − F(sx)^n)c_dδx^(δ−1).Source: B. Connectivity with retransmissions · p. 13
  • In the standard network, the expected number of nodes reached grows with log n, while choosing s_k = s_1/k makes it increase linearly with transmissions.The logarithmic growth follows from the digamma function in the fixed-threshold expression.Source: B. Connectivity with retransmissions · p. 13

IV. BROADCASTING

Broadcast reliability is characterized through the connection probability p_m(˜s), whose behavior approaches the no-fading limit as m increases. Taylor bounds yield sufficient conditions for high-probability connectivity, with tightness depending on the approximation and fading regime.

A. Broadcasting reliability

  • For every ˜s>0, p_m(˜s) increases with m and converges uniformly to its no-fading limit.The limit is 1 for ˜s<1 and 1/˜s for ˜s>1.Source: A. Broadcasting reliability · pp. 13–14
  • As m increases, p_m(˜s) approaches min{1, ˜s^-1}, the connection probability expected without fading.Without fading, nodes connect within distance 1/s; for ˜s>1, the fraction of nodes in [0,a] within that range is 1/˜s.Source: A. Broadcasting reliability · p. 14
  • The m-th order Taylor expansion at ˜s = 0 is a lower bound, while truncating the polynomial gives upper bounds.The passage identifies 1 + (m − 1)˜s as a natural first-order upper-bound choice.Source: A. Broadcasting reliability · p. 14
  • For m → ∞, bound (36) is not tight: its right-hand side tends to 1/e for positive ǫ, while the exact condition is as < 1/(1 − ǫ).Source: A. Broadcasting reliability · p. 15
  • The sufficient condition (37) is tight within 7% for ǫ < 0.1, and p_1(as) > 1 − ǫ can be solved exactly using the Lambert W function.Source: A. Broadcasting reliability · p. 15
  • The quadratic expansion gives the sufficient condition as < 2ǫ + 4/3ǫ^2, within 3.9% for ǫ < 0.25.The linear approximation yields the same bound as before.Source: A. Broadcasting reliability · p. 15

B. Broadcast transport sum-distance and capacity

Broadcast transport capacity depends on the path-loss parameter ∆ and transmission rate, while the sum-distance captures the spatial contribution of fading. The optimal rate choices are independent of Nakagami parameter m, and Rayleigh fading can reduce capacity at small ∆.

B. Broadcast transport sum-distance and capacity

C. Optimum broadcasting (superposition coding)

The superposition-coding capacity analysis uses a lower bound based on a rate corresponding to SNR. For ∆<1, the bound is significantly larger than the single-rate-decoding case; at and above ∆=1, the bound is unbounded, although at ∆=1 the actual capacity remains finite.

C. Optimum broadcasting (superposition coding)

V. OTHER APPLICATIONS

The other-applications section considers the average maximum transmission distance and bounds it using the distribution of connected nodes. The stated bound is tight in the illustrated setting, and Rayleigh fading's distance gain grows as the threshold decreases.

A. Maximum transmission distance

B. Probabilistic progress

Probabilistic progress balances transmission distance against success probability, while retransmission results characterize which nodes receive packets and support localization. The analysis finds that fading's effect depends on whether range or reliability is prioritized.

B. Probabilistic progress

  • Continuous probabilistic progress maximizes x^(1/α)P[f > sx]; for the standard network with α = 2, its optimum is x = 1/(2s).The objective is the product of distance-rate scaling and success probability.Source: B. Probabilistic progress · p. 20
  • Fig. 7 reports expected maximum transmission distances for the standard two-dimensional network over s ∈ [0.05, 1.00], and displays the non-fading-case comparison curve s^-1/2.Source: Figure 7 · p. 21, B. Probabilistic progress · p. 21
  • For discrete probabilistic progress, i_opt = 1 at large s, while the continuous optimizer is approximately cd/(αs) at small s.The continuous optimizer is used to guide rounding to a node index.Source: B. Probabilistic progress · p. 21
  • When reliability matters in the non-opportunistic setting, Rayleigh fading reduces transmission range by a factor α^(-1/α).The expected distance to the selected node is approximately (1/(αs))^(1/α).Source: B. Probabilistic progress · p. 21

C. Retransmissions and localization

  • The expected number of nodes receiving k packets is independent of n, the number of transmissions.The stated independence applies to the mean number of nodes receiving k packets.Source: C. Retransmissions and localization · p. 22
  • For the standard network with α = 2, Fig. 8 plots densities for nodes receiving exactly k of six messages; its dashed curve represents nodes receiving at least one packet.The plotted densities are normalized by the expected number of nodes receiving k packets.Source: C. Retransmissions and localization · pp. 22–23, Figure 8 · p. 23
  • The retransmission expression evaluates each additional transmission's contribution to broadcast transport sum-distance and capacity.The paper states that this expression permits the contribution to be evaluated.Source: C. Retransmissions and localization · p. 22
  • When path loss x is measured, node index can be estimated by maximum likelihood; in the standard network, the estimate is ˆi(x) = ⌈cd/x⌉.The estimate is tied to the standard-network setting.Source: C. Retransmissions and localization · pp. 22–23

VI. CONCLUDING REMARKS

The paper presents a point-process framework that incorporates fading to analyze wireless-network geometry. Its applications yield results about fading gains, path-loss and dimensionality, and broadcast transport capacity.

VI. CONCLUDING REMARKS

  • A point-process model incorporating both geometry and fading provides a geometric interpretation and a convenient route to closed-form wireless-network results.The framework is used for analytical investigations of network properties and fading's impact.Source: §VI · p. 23
  • For Nakagami-m fading, connectivity gain is the δ-th fading moment, while broadcast transport sum-distance gain is the ∆-th moment.The conclusion also reports negative fading impact when the path-loss exponent exceeds d for connectivity or d+1 for broadcasting.Source: §VI · p. 23
  • Broadcast transport capacity is unbounded when ∆ > 1, a result the authors say raises doubts about transport capacity as a performance metric.The conclusion identifies maximum transmission distance, probabilistic progress, retransmissions, and localization as other applications.Source: §VI · pp. 23–24
  • Network behavior depends critically on the ratio of dimension count to path-loss exponent, reflected in the recurring parameters δ and ∆.Source: §VI · p. 24
Loading 0711.1890v1…