PaperlayerView exact revision on arXiv

Source-linked AI summary

Cellular-Base-Station Assisted Device-to-Device Communications in TV White Space

Guoru Ding, Jinlong Wang, Qihui Wu, Yu-Dong Yao, Fei Song, Theodoros A. Tsiftsis

arXiv:1506.01394v1cs.NI

TL;DR

The paper addresses how to use TVWS for D2D communications while protecting licensed DTV services, by building a location-specific database with cellular infrastructure. It combines crowd measurements, matrix completion, boundary learning, and opportunistic power control, and simulations show improved reuse with reduced interference. The approach also targets TVWS created by location-dependent shadowing inside conventional DTV protection regions.

  • Problem

    The problem is enabling D2D communications in TVWS while satisfying strict interference constraints for licensed DTV services.

  • Method

    The method uses cellular base stations to collect crowd-sensed measurements, recover missing data, detect irregular DTV boundaries, and determine each D2D link’s MPEP.

  • Results

    The proposed approach improves spatial reuse and reduces interference to potential DTV receptions compared with the state-of-the-art approach.

  • Takeaways & Limitations

    The paper demonstrates that location-dependent shadowing can expose additional TVWS opportunities for small-scale D2D communications near large-scale DTV services.

Abstract

from arXiv · show

This paper presents a systematic approach to exploit TV white space (TVWS) for device-to-device (D2D) communications with the aid of the existing cellular infrastructure. The goal is to build a location-specific TVWS database, which provides a look-up table service for any D2D link to determine its maximum permitted emission power (MPEP) in an unlicensed digital TV (DTV) band. To achieve this goal, the idea of mobile crowd sensing is firstly introduced to collect active spectrum measurements from massive personal mobile devices. Considering the incompleteness of crowd measurements, we formulate the problem of unknown measurements recovery as a matrix completion problem and apply a powerful fixed point continuation algorithm to reconstruct the unknown elements from the known elements. By joint exploitation of the big spectrum data in its vicinity, each cellular base station further implements a nonlinear support vector machine algorithm to perform irregular coverage boundary detection of a licensed DTV transmitter. With the knowledge of the detected coverage boundary, an opportunistic spatial reuse algorithm is developed for each D2D link to determine its MPEP. Simulation results show that the proposed approach can successfully enable D2D communications in TVWS while satisfying the interference constraint from the licensed DTV services. In addition, to our best knowledge, this is the first try to explore and exploit TVWS inside the DTV protection region resulted from the shadowing effect. Potential application scenarios include communications between internet of vehicles in the underground parking, D2D communications in hotspots such as subway, game stadiums, and airports, etc.

I. INTRODUCTION

The paper proposes using mobile crowd sensing and cellular base stations to build localized TVWS databases for D2D links. It combines spectrum-data recovery, irregular DTV boundary detection, and opportunistic power control to improve reuse while limiting interference.

  • Motivation: TVWS offers unlicensed spectrum for D2D communications, but deployments must protect licensed DTV services from harmful interference.TVWS consists of locally unused TV bands, while spectrum sensing can suffer from hidden nodes caused by shadowing.
  • System goal: The proposed system builds a location-specific database that provides each D2D link with its maximum permitted emission power in an unlicensed DTV band.Cellular base stations support the database service for nearby D2D links.
  • Design rationale: Location-dependent shadowing creates coverage holes near or inside traditional DTV protection regions, providing additional opportunities for spatial spectrum reuse.The paper treats shadowing as a source of isolation between small-scale D2D links and large-scale DTV services.
  • Optimization and evaluation: The spatial-reuse problem maximizes unlicensed transmit power subject to device peak-power and licensed-reception interference constraints.The approach is evaluated in representative simulations against a state-of-the-art approach.
  • Approach: Mobile crowd sensing collects massive spectrum measurements, while matrix completion, support vector machines, and spatial reuse determine local TVWS availability and D2D power.The cellular base station collects, preprocesses, and mines measurements from its own and neighboring cells.
  • Network model: The network model combines a large-footprint DTV transmitter with interconnected cellular cells containing base stations and mobile devices.The DTV coverage boundary is irregular, while simplified protection and no-talk boundaries are represented separately.

B. Signal Propagation Model

The propagation model represents received TV signals through distance-dependent path loss and location-dependent shadowing. This modeling choice supports irregular DTV coverage boundaries and identifies TVWS opportunities that disc-based models can miss.

  • Propagation components: The channel model separates large-scale path loss, medium-scale shadowing, and small-scale fading.For wideband TV channels, small-scale fading is generally averaged out, leaving path loss and shadowing as the main modeled components.
  • Hybrid model: The adopted hybrid model uses a deterministic distance- and frequency-dependent path-loss component.The path-loss exponent depends on the propagation environment, while distance is measured in kilometers and frequency in megahertz.
  • Hybrid model: Shadowing is modeled as a Gaussian variable whose average loss depends on the environment or location.The mean shadowing loss can vary from 0 dB to tens of dBs.
  • Coverage definition: Coverage probability is defined from the minimum received power required for reliable DTV decoding and the average received power at a location.The average received power accounts for transmit power, path loss, and average shadowing loss.
  • Coverage definition: The DTV coverage area and TVWS are defined as, respectively, locations meeting the coverage threshold and locations outside that coverage set.The coverage threshold is denoted by ν_cov.
  • Implications: Location-dependent shadowing produces irregular ground-truth coverage boundaries and can reveal exploitable TVWS near or inside conventional disc coverage regions.Disc models that neglect or spatially average shadowing cannot represent these local coverage variations.

D. Unlicensed Device Interference Model

The paper models each unlicensed device’s interference area and maximizes its transmit power while protecting licensed DTV receptions. The formulation accounts for irregular coverage boundaries, heterogeneous spatial reuse, hardware limits, and interference-probability constraints.

  • Interference model: The interference area contains locations where an unlicensed device can exceed the interference tolerance threshold I_max for DTV reception.The model defines a location interference threshold ν_int and expresses the corresponding interference threshold using path loss and shadowing terms.
  • Interference model: Because DTV receiver locations are unknown, the model conservatively assumes active receivers may occur anywhere inside the DTV coverage area.A no-talk region combines the DTV protection region with an additional keep-out distance dependent on interference range.
  • Problem formulation: The objective is to maximize an unlicensed device’s transmit power P*_xi for a given location while satisfying hardware and interference-probability constraints.The peak transmit power is bounded by device hardware, and ν_int protects licensed DTV receptions.
  • Problem formulation: The worst-case DTV receiver position is the boundary location that perceives the strongest interference from a device and supports rewriting OP1.The resulting optimal powers vary across device locations.
  • Problem formulation: Irregular DTV boundaries make OP1 mathematically difficult, while limited measurements and strict interference constraints complicate local radio-environment estimation.Collaborative measurement sharing can help, but its energy, delay, and deployment costs may be commercially unwelcome.
  • Problem formulation: Spatial reuse opportunities are heterogeneous because devices at different locations, including neighboring devices, may have different MPEPs.This heterogeneity makes a single global solution infeasible.

IV. OVERVIEW OF THE PROPOSED SOLUTION

The proposed solution builds a location-specific TVWS database using cellular infrastructure and active spectrum measurements. It sequentially reconstructs missing data, detects irregular DTV boundaries, and enables opportunistic D2D spectrum sharing.

  • Key Ideas: A cellular base station builds a location-specific TVWS database from active spectrum measurements collected with existing cellular infrastructure.The database provides localized spectrum service for mobile devices and can be updated offline and asynchronously because DTV coverage is relatively static.
  • Building blocks: Mobile crowd sensing collects location-aware spectrum measurements from massive personal mobile devices.The approach uses the devices’ sensing capabilities and positioning information to gather local spectrum data.
  • Building blocks: Matrix completion recovers missing spectrum measurements from known data through spatial interpolation.This addresses locations lacking sufficient measurements.
  • Building blocks: Boundary detection mines completed spectrum data to identify the irregular coverage boundary of a licensed DTV transmitter.The detected boundary supports subsequent spatial-reuse decisions.
  • Building blocks: Spatial reuse allows unlicensed D2D links to opportunistically share spectrum with licensed DTV services.The sequence is implemented at each cellular base station without modifying existing infrastructure.

V. ALGORITHM DESIGNS OF THE PROPOSED SOLUTION

The algorithm design uses mobile crowd sensing to gather localized measurements, cleanses or identifies unreliable reports, and reconstructs missing spectrum data for downstream boundary analysis. These components support cellular-base-station processing of TVWS information.

  • Mobile crowd sensing: Mobile crowd sensing lets smartphones, tablets, and in-vehicle sensors contribute small amounts of spectrum data to form a large dataset.The method is intended to replace specialized deployed sensors with measurements gathered from personal devices.
  • Mobile crowd sensing: Each mobile device senses DTV signal strength with a mobile app and reports it with its current location to its cellular base station.The base station stores the measurements in a local operational database.
  • Measurement quality: Crowd measurements may be unreliable, untrustworthy, or malicious, so the sensing model includes an abnormal-data component.The abnormal component is zero for normal data and nonzero for abnormal data.
  • Measurement quality: Prior data-cleansing and unsupervised-clustering methods can remove abnormal sensing components and distinguish spectrum attackers from reliable sensors.The methods use corrupted-data cleansing and historical-report dissimilarity analysis, respectively.

2) Incentive to Contribute:

Mobile crowd sensing creates a localized TVWS database but requires incentives for participation and methods to handle collection overhead and incomplete spatial measurements. The paper uses matrix completion with FPCA to reconstruct missing spectrum data, with recovery improving as sampling becomes denser and spatial resolution increases.

  • Incentive to Contribute:: Participating devices consume battery and computing resources and face potential privacy threats from sharing spectrum data with location information.The paper suggests rewards such as bandwidth or discounted or free communication minutes when TVWS is available.
  • Spectrum Data at Each Cellular BS:: Cellular base stations collect, preprocess, and mine spectrum measurements from personal devices to form localized TVWS database services for nearby D2D links.The collected data are multidimensional measurements of received DTV signal power across time, frequency, and space.
  • Uplink Overhead Analysis:: Mobile crowd sensing adds uplink overhead because devices must send spectrum measurements to the cellular base station.The paper proposes reducing overhead by increasing the collection period and participating-device count while decreasing the number of measurements and their bit length.
  • Matrix Completion for Unknown Measurements Recovery: Incomplete crowd measurements arise because human activity leaves some locations with few or unreliable samples, motivating matrix completion instead of explicit spatial-weight design.The cellular base station represents the spectrum dataset as a partially observed matrix and recovers unobserved entries through nuclear-norm minimization solved with FPCA.
  • Matrix Completion for Unknown Measurements Recovery: -20 dB recovery RSE is approached at 80×80 m spatial resolution when the sampling rate is at least 30%.Recovery error generally decreases with more samples per grid and higher sampling rate, while smaller grids yield better recovery performance.

C. Support Vector Machine for Coverage Boundary Detection

The cellular BS reconstructs noisy, incomplete spectrum declarations and uses SVM classifiers to detect irregular DTV coverage boundaries for spatial reuse. Detection improves with finer grids and higher sampling, while RBF SVM outperforms quadratic SVM.

  • Coverage declarations: The cellular BS converts recovered DTV signal strengths at grid locations into binary coverage declarations for boundary detection.The threshold includes an offset δP that can conservatively protect licensed DTV services or aggressively improve reuse.
  • SVM formulation: Because declarations may be erroneous and their correctness is unknown, boundary detection minimizes classification errors using SVMs.The formulation begins with linear SVM and extends to nonlinear classifiers through kernelization.
  • Nonlinear boundary detection: Kernelized SVMs accommodate nonlinear and irregular DTV boundaries caused by attenuation from hills or buildings.Polynomial and RBF Gaussian kernels map the original two-dimensional data into a higher-dimensional feature space where separation is generally easier.
  • Performance: Smaller grid sizes, higher sampling rates, and RBF Gaussian SVM generally yield better successful boundary-detection probability.The probability measures correct inside/outside classification relative to the ground-truth DTV coverage area.
  • Computational complexity: SVM training has dataset-dependent quadratic-programming complexity between O(N^2) and O(N^3), where N is the number of training samples.This complexity is the main computational constraint identified for the SVM component.

D. Opportunistic Spatial Reuse between a D2D Communication Link and the DTV Services

After detecting the DTV boundary, each cellular BS assigns a maximum permitted emission power to D2D links according to whether their locations are black, complete white, or gray space. These assignments support localized TVWS database lookups.

  • MPEP assignment: The opportunistic reuse algorithm computes MPEP for any device location inside the cellular BS coverage area using the detected DTV boundary.The resulting values can be stored for all cell locations and returned through a location-based database lookup.
  • Black space: A grid covered by DTV is black space, so D2D transmission is prohibited to protect potential DTV receptions.This is the no-transmission case of the spatial reuse policy.
  • Complete white space: A grid outside DTV coverage with no boundary intersection in the device’s worst-case interference area is complete white space, allowing peak transmit power Ppeak.The policy treats the link as unable to cause harmful interference under the specified boundary condition.
  • Gray space: A grid outside DTV coverage whose worst-case interference area intersects the boundary is gray space, allowing only an intermediate power P*_xi ∈ (0, Ppeak).The selected power is constrained by the strongest-interference boundary location, called the WCRP.

A. Simulation Setup

The simulations evaluate the proposed TVWS reuse approach in two challenging settings where a cellular cell partly overlaps the DTV coverage region. Results show that crowd sensing, matrix completion, and SVM boundary detection recover usable coverage information and reduce reuse errors relative to a model-based approach.

  • Simulation scenarios: The simulations focus on cells partially inside and partially outside DTV coverage, the technically most challenging deployment case.Scenario I places the BS at the DTV protection-region edge; Scenario II places it inside the region with shadowing-created white space.
  • Scenario I setup: Scenario I uses an 8 km × 8 km area divided into 100 × 100 grids, with the cellular BS centered in the area.The ground-truth boundary is nearly linear because the cell lies at the edge of DTV coverage.
  • Scenario I pipeline: With only 50% of locations sampled, crowd sensing produces incomplete noisy data, matrix completion makes the coverage recognizable, and SVM further derives the boundary.The detected boundary appears to perform well when compared with the ground-truth coverage.
  • Comparison with model-based approach: The traditional model-based approach yields MPEP biases from -0.08 to 0.1 and interference-probability biases from -0.1 to 0.9, depending on localization error.The cited setup considers maximum localization errors of 50 m for GPS, 150 m for Wi-Fi, and 1000 m for GSM-based localization.
  • Results: The proposed approach has smaller MPEP and interference-probability bias ranges, and δP can make both maximum biases below 0 while keeping interference probability below νint = 0.1.The reported advantages are improved spatial reuse and reduced interference near the ground-truth coverage edge.

C. Simulation Results under Scenario II

Scenario II evaluates TVWS reuse inside a DTV protection region with irregular shadowing-induced coverage. The proposed approach enables D2D transmission while meeting DTV interference constraints at most cell locations.

  • Scenario II setting: Scenario II places a cellular base station inside the DTV protection region, where shadowing creates a more irregular TVWS coverage boundary.The scenario is more complex than Scenario I because TVWS lies inside the protection region.
  • Comparison with the model-based approach: The model-based approach derives zero MPEPs for every localization-error setup, so no D2D transmission is allowed.Its interference-probability bias is -0.1 under ν_int = 0.1.
  • Proposed approach: More than 90% of cell locations satisfy the licensed-DTV interference constraint under the proposed approach for all three δP settings.The approach therefore enables D2D communications in TVWS inside the protection region.
  • Parameter tradeoff: Larger δP reduces interference to potential DTV receptions but increases MPEP bias, creating a protection–reuse tradeoff.The study identifies δP = 3 as a better balance for this scenario.
  • Implications: The paper presents exploiting TVWS inside a DTV protection region as a first attempt, with possible use in underground parking and hotspot environments.Examples include internet-of-vehicles communications and D2D communications in subways, stadiums, and airports.
  • System approach: The overall system combines crowd sensing, matrix completion, nonlinear support vector machines, and opportunistic spatial reuse to determine each D2D link’s MPEP.The cellular infrastructure supports the location-specific TVWS database and its spatial-reuse decisions.

APPENDIX A MATRIX SHRINKAGE OPERATOR Sν(·)

The matrix shrinkage operator applies singular-value thresholding to a matrix representation. It reduces each nonnegative singular value by ν and sets values below ν to zero.

  • Operator definition: For M = Udiag(σ)V^T, the operator S_ν(·) acts on the singular values in the matrix’s singular value decomposition.The matrix has dimensions p × m, with U, σ, and V defined through the decomposition.
  • Shrinkage behavior: S_ν reduces every nonnegative singular value by ν, truncating any singular value smaller than ν to zero.This is the matrix shrinkage rule used by the operator.
Loading 1506.01394v1…