PaperlayerView exact revision on arXiv

Source-linked AI summary

Transmission Capacity of Ad Hoc Networks with Spatial Diversity

Andrew M. Hunter, Jeffrey G. Andrews, Steven Weber

arXiv:cs/0702044v2cs.IT

TL;DR

The paper asks which multiple-antenna techniques most improve capacity in random ad hoc networks, where prior capacity work emphasized network-size scaling. It derives outage probabilities and transmission capacity for several single-datastream diversity and beamforming methods, finding that beamforming provides the strongest gains while space-time coding offers marginal improvement.

  • Problem

    The paper examines how capacity scales with antennas and which multiple-antenna technique can support the densest ad hoc network.

  • Method

    The paper analyzes outage probabilities and transmission capacity for random networks using sectorized antennas, eigen-beamforming, space-time block coding, and selection combining.

  • Results

    Static beamforming has λ_ϵ = Θ(M_tM_r) with ideal sectorized antennas, while maximal-ratio transmission and combining has λ_ϵ = O((M_tM_r)^(2/α)).

  • Takeaways & Limitations

    Beamforming provides the largest transmission-capacity increases; selection combining has rapidly diminishing returns, and space-time block coding yields little gain beyond two antennas.

Abstract

from arXiv · show

This paper derives the outage probability and transmission capacity of ad hoc wireless networks with nodes employing multiple antenna diversity techniques, for a general class of signal distributions. This analysis allows system performance to be quantified for fading or non-fading environments. The transmission capacity is given for interference-limited uniformly random networks on the entire plane with path loss exponent $α>2$ in which nodes use: (1) static beamforming through $M$ sectorized antennas, for which the increase in transmission capacity is shown to be $Θ(M^2)$ if the antennas are without sidelobes, but less in the event of a nonzero sidelobe level; (2) dynamic eigen-beamforming (maximal ratio transmission/combining), in which the increase is shown to be $Θ(M^{\frac{2}α})$; (3) various transmit antenna selection and receive antenna selection combining schemes, which give appreciable but rapidly diminishing gains; and (4) orthogonal space-time block coding, for which there is only a small gain due to channel hardening, equivalent to Nakagami-$m$ fading for increasing $m$. It is concluded that in ad hoc networks, static and dynamic beamforming perform best, selection combining performs well but with rapidly diminishing returns with added antennas, and that space-time block coding offers only marginal gains.

I. INTRODUCTION

The paper asks how multiple antennas change ad hoc network capacity and which diversity or beam-steering technique supports the densest network. It develops transmission-capacity comparisons for single-datastream techniques under random-network conditions.

  • Motivation: The paper focuses on how capacity scales with antennas at each node and which technique supports the densest network.This differs from prior work focused mainly on capacity as network size grows.
  • Scope: MIMO techniques are grouped into diversity-achieving, beam-steering, and spatial multiplexing approaches.The paper concentrates on diversity and beam-steering because they preserve a single datastream and are easier to compare fairly.
  • Metric: Transmission capacity is the maximum successful-transmission density multiplied by data rate under an outage constraint.For rate b, outage ϵ, and optimal contention density λ_ϵ, c_ϵ = b(1 − ϵ)λ_ϵ.
  • Scope: The analysis compares beamforming, antenna sectorization, space-time block coding, and selection combining in single-datastream ad hoc networks.Spatial multiplexing is omitted as future work, although some developed results may apply to it.
  • Main scaling results: For ideal sectorized antennas, λ_ϵ = Θ(M_tM_r), while maximal-ratio transmission and combining achieve λ_ϵ = O((M_tM_r)^(2/α)).These are small-outage scaling relations for transmit and receive antenna counts with α > 2.
  • Main conclusions: Beamforming achieves the largest transmission-capacity increases, whereas space-time block coding yields little gain beyond two antennas.Selection and receiver diversity also matter, with receiver techniques avoiding feedback requirements in the stated comparison.

B. The Optimal Contention Density

The paper derives exact outage probabilities and low-outage optimal contention densities for random-access networks, using a general interference model and broad received-signal distributions. The framework separates received-signal effects from interfering-signal statistics while imposing explicit modeling conditions.

  • Theorem: For small outage constraints, solving the outage condition yields the optimal contention density λ_ϵ.The analysis uses λ_ϵ = λ̄_ϵ + O(ϵ^2), with the error term implicitly suppressed.
  • Conditions and interpretation: The framework applies to finite sets of received-signal and fading distributions when the desired signal is independent of aggregate interference.This includes several MIMO techniques and separates K_α, determined by received-signal distributions, from C_α, determined by interfering-signal statistics.
  • Theorem: The theorem gives exact outage probability for any network density or target SIR and an optimal contention density in the low-outage regime.Thermal noise is omitted for convenience but can be reinserted.
  • Operating assumptions: The analysis assumes fixed outage constraint, data rate, and transmitter-receiver distance for clarity.If communicating pairs use longer distances at a given interferer density, outage probability or data rate must suffer.
  • Distance variation: Variable transmission distances reduce transmission capacity by E[R^2]/E[R]^2, while the relative benefits of the MIMO techniques remain unaffected.The small-outage approximation must remain valid over the permitted distances.

III. TRANSMISSION CAPACITY IN LOS AND NLOS ENVIRONMENTS

The paper extends transmission-capacity analysis from fading to non-fading environments using Nakagami-m fading, with Rayleigh fading as m = 1 and non-fading approached as m increases. It shows that fading severity and path loss affect capacity gains, motivating diversity techniques that mitigate fading.

  • Nakagami-m fading includes Rayleigh fading at m = 1 and approaches non-fading as m increases.The model uses integer m to characterize environments between Rayleigh fading and non-fading propagation.
  • The optimal contention density is derived for random-access single-antenna networks under Nakagami-m fading.The result applies to m ∈ N and an outage constraint ǫ.
  • Kα,m and Cα,m both increase as Θ(m ...), linking stronger channel conditions to higher capacity-related quantities.The supplied passage truncates the exponent, so the asymptotic expression is preserved only to the stated extent.
  • The Nakagami-m result bridges fading and non-fading environments and demonstrates potentially significant network-capacity gains relative to non-fading environments.The paper also notes greater improvement from strong line-of-sight conditions when path loss is lower.
  • Lower path loss suffers more from severe fading when β > 1 and improves more with strong line-of-sight propagation.This distinction is particularly relevant to dense networks communicating with nearby neighbors.

IV. SECTORIZED ANTENNAS

The paper analyzes sectorized directional antennas with configurable sidelobes and derives their interference and transmission-capacity effects. Sectorization can nearly provide a quadratic gain in the number of antennas at low sidelobe levels, but practical sidelobes and propagation assumptions constrain that gain.

  • Sectorized antenna model: Sectorized antennas use M directional sectors, with each antenna covering 2π/M radians and providing aperture gain M within its sector.The model also permits a constant fractional sidelobe level γ for out-of-sector transmitted and received power.
  • Interference model: The interference model separates interferers into four processes according to sector membership and transmit direction.Their effective densities, angular regions, and combined antenna gains determine the corresponding interference contributions.
  • Capacity scaling: γ^-4α is an upper bound on the transmission-capacity increase due to antenna sectorization.The bound follows from the constant sidelobe-level model.
  • Capacity scaling: Nearly M^2 transmission-capacity improvement occurs with low sidelobe levels.The result identifies interference reduction as the main physical-layer benefit of directional antennas.
  • Limitations: Nonzero sidelobes limit gains even for highly directional antennas, while realistic multipath can make static antennas less robust.The analysis assumes fading statistics remain unchanged under sectorization, an assumption that may fail for very directional antennas.

V. TRANSMISSION CAPACITY OF EIGEN-BEAMFORMING NETWORKS

The paper analyzes eigen-beamforming in interference-limited ad hoc networks, where channel-aware weights maximize desired-signal power without knowledge of interfering channels. For 1 × M combining, antenna gains improve contention density toward an upper bound, while path loss governs the balance between spatial separation and interference robustness.

  • Eigen-beamforming: Dynamic beamforming uses dominant channel singular vectors or eigenvectors to weight transmitted or received signals coherently.The analysis assumes perfect knowledge of each node’s own channel but not of interfering channels.
  • Eigen-beamforming: Maximal ratio combining with M receive antennas is equivalent to maximal ratio transmission with M transmit antennas and one receive antenna.The proposition treats these configurations as equivalent under Rayleigh fading.
  • Analysis: The optimal contention density is characterized under Rayleigh fading for small outage constraints, with K_α,M and C_α determining the result.C_α equals the single-antenna factor C_α,1, while K_α,M varies with antenna count.
  • Analysis: For M-antenna combining, the desired received-signal power follows a chi-square distribution with 2M degrees of freedom, while interference retains the single-antenna shot-noise form.This separates the diversity effect on the desired signal from the interference model.
  • Scaling behavior: Higher path loss improves transmission capacity at smaller antenna counts, whereas larger arrays increasingly benefit from interference robustness rather than spatial separation.As path loss decreases, the gain of MIMO over SISO increases, but higher path loss can initially separate transmissions more effectively.

B. Mt × Mr MIMO Eigen-Beamforming

Dynamic eigen-beamforming uses the dominant singular vectors of the desired MIMO channel, yielding bounds on outage and optimal contention density. Its gains diminish as path loss increases, and explicit expressions are unavailable for arbitrary antenna configurations.

  • Method: Mt × Mr eigen-beamforming increases diversity order to MtMr by beamforming along the maximum singular vectors of the desired channel.The received desired-signal power equals the square of the maximum singular value.
  • Scaling with path loss: Lower path loss produces greater gains over SISO, while higher path loss makes spatial diversity less beneficial as nodes become separated through attenuation.The factor approaches one as α increases.
  • Contention-density result: Proposition 4 bounds the optimal contention density for random-access networks using maximal ratio transmission and combining under small outage constraints.The bound applies to Mt transmit and Mr receive antennas.
  • Interference analysis: The interference distribution remains unchanged from the single-antenna Rayleigh-fading case, giving Cα = Cα,1.Independence between the receive beamformer, interfering channel, and interferer beamformer enables the simplification.
  • Bounds and limitations: An explicit expression for Kmrt,α,Mt,Mr is unavailable for arbitrary Mt and Mr, so the analysis uses bounds and a specific 2 × 2 example.The paper discusses lower and upper bounds for larger antenna systems and different propagation conditions.

VI. TRANSMISSION CAPACITY OF OSTBC NETWORKS

The OSTBC analysis derives optimal contention density for transmit block coding with receive MRC. Although coding hardens the desired channel, repeated symbols also amplify interference, leaving little network-capacity gain.

  • Motivation and code structure: OSTBCs provide full point-to-point diversity without transmitter channel-state information and use a matched-filter receiver.The Alamouti code is given as the familiar example with two transmit antennas and two time slots.
  • Model and result: Proposition 5 gives the optimal contention density for OSTBC transmitters with Mt antennas and code parameter Nr, combined with Mr-antenna MRC receivers.The result assumes Rayleigh fading and an outage constraint ε.
  • Signal distribution: The desired post-processing signal has a χ2 distribution with 2MtMr degrees of freedom, equivalent to MRC with M = MtMr.This lets the MRC-based theorem determine the OSTBC K factor.
  • Interference analysis: OSTBC interference analysis assumes rough independence between desired and interfering statistics because repeated interference terms are not strictly independent.The assumption makes the analysis worst case.
  • Conclusion: OSTBCs offer marginal gains in significant cochannel interference: channel hardening helps, but repeated symbols tend to amplify interference.The resulting gain is comparable to reducing fading through Nakagami-m channel hardening.
  • Capacity comparison: Beyond two antennas, OSTBC transmission capacity can fall below MRC because reduced-rate codes offset contention-density gains; with one receive antenna, additional transmit antennas provide essentially no gain.The paper identifies receive diversity as the primary source of gain.

VII. TRANSMISSION CAPACITY OF SELECTION DIVERSITY AND COMBINING NETWORKS

Selection diversity improves ad hoc network performance by choosing favorable antenna pairs without amplifying interference. The gains are appreciable but diminish as the antenna array grows.

  • Selection schemes: Selection diversity chooses the strongest available antenna or antenna pair, offering simpler implementation than more sophisticated combining schemes.The best-pair scheme selects one transmit and one receive antenna from the full channel matrix.
  • Analytical result: Proposition 6 characterizes optimal contention density when nodes select the best transmit-receive antenna pair among Mt and Mr antennas.The interference statistics remain identical to the SISO case for any selected pair.
  • Capacity gains: Antenna selection significantly enhances network performance because it improves the desired channel without amplifying interference.The paper compares receive-only selection with joint transmit-receive best-pair selection.
  • Diminishing returns: As antenna counts increase, obtaining M^2 statistically independent antenna pairs becomes difficult, and array gain quickly becomes superior.Selection can therefore trade performance against implementation complexity.

VIII. CONCLUSION

The paper quantifies outage probability and transmission capacity for several spatial-diversity techniques in random ad hoc networks. Beamforming performs best, selection combining provides appreciable gains, and OSTBC offers marginal gains.

  • Contributions: The analysis derives exact outage probabilities and small-outage optimal contention densities for a broad class of received-signal distributions.The distributions cover maximal ratio transmission/combining, OSTBC, selection diversity/combining, and static sectorized beamforming.
  • Main comparison: Beamforming and sectorized antennas produce the largest transmission-capacity improvements among the evaluated diversity techniques.Directional antennas approach a factor of M^2 at low sidelobe levels.
  • Main comparison: Selection combining provides appreciable gains, but its returns diminish as the number of antennas increases.The gain is higher when interference is more severe.
  • Main comparison: OSTBC systems provide marginal gains at best because channel hardening is offset by interference effects from repeated symbols.Their gains are equivalent to those from reduced fading in Nakagami-m channels.
  • Interference and path loss: Beamforming and selection gains increase under more severe interference and lower path loss, whereas OSTBC behaves oppositely.The comparison concerns transmission-capacity gains in random ad hoc networks.

APPENDIX I PROOF OF THEOREM 1

The proof derives transmission success and outage probabilities through Laplace-transform representations, then solves for the optimal contention density. It also extends the expansion to include thermal noise under a stated validity condition.

  • The probability of successful transmission is expressed using the Laplace transform of the interference distribution.
  • Derivatives of the interference Laplace transform and combinatorial subset expansions support the derivation of the general expression.
  • First-order Taylor expansions discard higher-order terms in κ and produce the outage-probability expression.
  • Solving with ζ = βR^α and K_α yields the transmission-capacity result through the optimal contention density.
  • Thermal noise is incorporated by replacing interference with interference plus noise, whose Laplace transform multiplies the interference transform.
  • The noise-inclusive expansion is valid only when interference-free outage from intended-signal fading and thermal noise is below ε.

APPENDIX II PROOF OF PROPOSITION 1

The proof specializes the general framework to Nakagami fading with distinct desired and interfering parameters and analyzes the resulting contention-density bounds. It then characterizes the approach to the non-fading limit as the fading parameter increases.

  • Nakagami fading with separate desired-signal and interferer parameters provides the specialized received-power complementary CDF.
  • The Nakagami mark distribution modifies the interference transform, and Theorem 1 gives the corresponding optimal contention density.
  • For Rayleigh desired-signal fading and increasingly non-fading interferers, the interferer fading mark transform approaches e^(-ζ|x|^-α).
  • With non-fading interferers and increasing desired-signal Nakagami parameter, the desired-signal distribution approaches an impulse at S_0 = 1.
  • The asymptotic order of K_α,m is determined for fixed C_α,∞ using bounds on the optimal contention density.
Loading cs/0702044v2…