PaperlayerView exact revision on arXiv

Source-linked AI summary

A Measurement Based Shadow Fading Model for Vehicle-to-Vehicle Network Simulations

Taimoor Abbas, Katrin Sjöberg, Johan Karedal, Fredrik Tufvesson

arXiv:1203.3370v5cs.NI

TL;DR

Existing V2V models often neglect shadowing from vehicles, despite its relevance to received power and interference. This paper develops a measurement-based LOS/OLOS model for urban and highway VANET simulations and compares it with a Nakagami-based model, observing about 10 dB of additional loss under vehicle-obstructed LOS.

  • Problem

    Vehicle-induced shadowing is largely neglected in V2V channel models, although distinguishing LOS, OLOS, and NLOS conditions is needed to characterize their distinct channel properties.

  • Method

    The paper uses real highway and urban measurements, video-based LOS/OLOS/NLOS separation, correlated log-normal shadowing, and a VANET-simulator implementation.

  • Results

    About 10 dB of additional received-power loss is observed when vehicles obstruct LOS, and simulations show differences between the LOS/OLOS and Nakagami-based models.

  • Takeaways & Limitations

    The LOS/OLOS model provides a measurement-based way to represent vehicle obstruction and correlated shadowing in VANET system simulations.

Abstract

from arXiv · show

The vehicle-to-vehicle (V2V) propagation channel has significant implications on the design and performance of novel communication protocols for vehicular ad hoc networks (VANETs). Extensive research efforts have been made to develop V2V channel models to be implemented in advanced VANET system simulators for performance evaluation. The impact of shadowing caused by other vehicles has, however, largely been neglected in most of the models, as well as in the system simulations. In this paper we present a shadow fading model targeting system simulations based on real measurements performed in urban and highway scenarios. The measurement data is separated into three categories, line-of-sight (LOS), obstructed line-of-sight (OLOS) by vehicles, and non line-of-sight due to buildings, with the help of video information recorded during the measurements. It is observed that vehicles obstructing the LOS induce an additional average attenuation of about 10 dB in the received signal power. An approach to incorporate the LOS/OLOS model into existing VANET simulators is also provided. Finally, system level VANET simulation results are presented, showing the difference between the LOS/OLOS model and a channel model based on Nakagami-m fading.

I. INTRODUCTION

V2V channel models must distinguish propagation conditions because vehicle and building obstructions alter received power and channel behavior. The paper introduces a measurement-based LOS/OLOS shadow-fading model for VANET simulations.

  • V2V channels differ from cellular channels because both endpoints are mobile, use an ad hoc topology, and place antennas near ground level.
  • Path loss, small-scale fading, and shadow fading are distinct phenomena that all contribute to V2V channel characterization.Obstacles can attenuate the LOS component and other major multipath components.
  • Vehicle and building obstructions can block LOS, making received power dependent on propagation environment and LOS availability.
  • Most prior models lump LOS and non-LOS samples together, an assumption described as unrealistic, especially at larger distances.
  • Ignoring vehicles as obstacles can produce unrealistic assumptions about received power, interference levels, and higher-layer V2V behavior.
  • The paper models measured highway and urban shadow fading by separating LOS and OLOS conditions and provides VANET-simulator integration guidance.Temporal shadow-fading correlation is modeled as an auto-regressive process, and simulations compare the model with Cheng’s Nakagami-based model.

II. METHODOLOGY

The measurement campaign used a MIMO channel sounder with roof-mounted antennas on two vehicles, while synchronized video and GPS supported propagation-condition labeling.

  • Measurements used the RUSK-LUND channel sounder for MIMO channel-transfer-function measurements at 5.6 GHz with 200 MHz bandwidth.The setup sampled 641 frequency points and used short-term and long-term measurement durations.
  • Two 1.47 m high Volvo V70 station wagons carried roof-mounted omnidirectional antennas during the campaign.
  • Synchronized video, GPS, and channel measurements were combined during post-processing to identify LOS, OLOS, and NLOS conditions.

B. Measurement routes

Measurements covered highway and urban routes with differing traffic and roadside environments. Video-based labeling captured repeated transitions among LOS, OLOS, and NLOS states during vehicle movement.

  • Measurement routes: Eight routes in highway and urban environments varied in traffic density, roadside surroundings, scatterers, pedestrians, and houses.All measurements took place around Lund and Malmö in southern Sweden.
  • Measurement routes: Highway measurements used convoy travel at 22−25 m/s on a two-lane-per-direction road with traffic varying from low to high.
  • Measurement routes: Urban measurements included convoy and opposing-direction travel at 0−14 m/s on one- or two-lane streets lined with 2−4-story buildings.
  • LOS, OLOS and NLOS separation: Video frames at 25 frames per second were evaluated frame by frame to classify samples as LOS, OLOS, or NLOS.LOS required one camera to see the middle of the other vehicle’s roof; otherwise the link was blocked and further categorized.
  • LOS, OLOS and NLOS separation: The TX−RX link transitioned from LOS to OLOS 61 times in urban measurements and 23 times on highways.LOS-to-NLOS transitions occurred 4 times in urban data and 0 times in highway data; no OLOS-to-NLOS transition occurred.

D. Pathloss Derivation

The pathloss derivation averages power-delay profiles to suppress small-scale fading, extracts channel gain, corrects implementation effects, and relates gain to TX−RX distance.

  • The time-varying power-delay profile is averaged over Navg samples to eliminate small-scale fading before pathloss estimation.
  • Noise-thresholded averaged power-delay profiles yield the averaged channel gain through their zeroth-order moment.Noise is estimated from large-delay regions without transmitted-signal contributions.
  • Antenna influence, cable attenuation, and low-noise-amplifier effects are removed before calculating distance-dependent path loss.
  • Figure 2 compares overall channel-gain behavior with LOS/OLOS-separated Gaussian histograms and log-normal CDF fits for highway and urban data.The displayed distance bins are 20.4−29.1 m for highway and 84.6−121 m for urban data.
  • GPS-derived TX−RX distances are interpolated with a cubic spline to match the channel snapshots.The GPS sampling rate was one position per second, requiring interpolation.

E. Large-scale or shadow fading

The paper removes small-scale fading and analyzes large-scale channel variations separately by LOS, OLOS, and NLOS conditions. Separated LOS and OLOS data fit log-normal distributions, with OLOS showing substantially lower mean gain and an assumption required for incomplete measurements.

  • Averaging received power over 15λ removes small-scale fading before modeling large-scale variations from buildings and vehicles.The resulting averaged envelope is modeled using a log-normal distribution.
  • Separating LOS, OLOS, and NLOS data is necessary because unsplit distance-bin distributions show excessive and inconsistent spread.The additional attenuation was associated with LOS obstruction.
  • About 10 dB separates the mean LOS and OLOS channel-gain distributions, while both separated data sets are modeled as log-normal.For highway and open scenarios, higher losses are expected for obstructed LOS at communication distances below 80 m.
  • OLOS samples can fall below the channel sounder's noise floor, so larger-distance bins are assumed to continue the log-normal distribution observed at shorter distances.The assumption is made because incomplete OLOS data cannot be detected correctly below the noise threshold.

III. CHANNEL MODEL

The channel model is designed for VANET system simulations, combining realistic shadowing effects with reasonable complexity. It also specifies how to represent temporal shadow-fading correlation.

  • The LOS/OLOS shadow-fading model targets VANET simulations that require realistic shadowing with reasonable complexity.

A. Pathloss Model

The pathloss model separates LOS, OLOS, and NLOS conditions, fits distance-dependent LOS/OLOS behavior to measurements, and supplements insufficient NLOS measurements with an external low-complexity model. The fitted LOS and OLOS curves differ by roughly 8.6–10 dB, while parallel-street NLOS losses are reported above 120 dB.

  • Pathloss Model: Pathloss parameters are extracted for LOS and OLOS, but insufficient measured data prevents direct modeling of NLOS pathloss.NLOS behavior is therefore derived from models targeting similar scenarios.
  • Pathloss Model: The measured highway and urban channel gains are plotted against TX-RX distance with least-square fits to the deterministic part of the pathloss model.The model uses a generic log-distance power law with distance, pathloss exponent, and correlated Gaussian shadowing.
  • Pathloss Model: The dual-slope model uses pathloss exponent n1 before breakpoint distance db and n2 afterward, with standard deviation σ in both regions.A piecewise-linear form is used because it represents measurement data more accurately in practice.
  • Pathloss Model: The model validity range is d > 10 m because close-range GPS distances are unreliable and few samples exist below 10 m.The reference distance is set to d0 = 10 m.
  • Pathloss Model: The fitted LOS and OLOS channel gains differ by an offset of about 8.6–10 dB, consistent with previously reported vehicle-obstruction attenuation.The cited prior observations include 9.6 dB and 10–20 dB attenuation ranges.
  • NLOS Model: For parallel streets blocked by buildings, reported pathloss exceeds 120 dB, so interference from such vehicles can be ignored.The intersecting-street NLOS model is not advisable for parallel streets.

B. Spatial Correlation of Shadow Fading

Shadow fading is treated as a spatially correlated process after removing the distance-dependent mean. Its autocorrelation is modeled with a negative exponential whose decorrelation distance is scenario-dependent and estimated separately for LOS and OLOS.

  • Shadowing persists while a vehicle remains in a shadow region, motivating analysis of spatial correlation and average decorrelation distances.The persistence reflects a spatially correlated process.
  • The distance-dependent mean is subtracted so the remaining shadow-fading process can be treated as stationary.
  • The autocorrelation is modeled using Gudmundson’s negative exponential function.
  • The decorrelation distance dc is defined where the autocorrelation equals 1/e and is estimated from LOS and OLOS measurements for highway and urban scenarios.Table III reports the scenario-specific decorrelation distances.

C. Extension in the Traffic Mobility Models

The simulator extension identifies whether vehicles or buildings obstruct the TX–RX path and then applies the corresponding propagation model. It uses geometric information and, where available, Fresnel-zone characteristics to distinguish LOS and OLOS conditions.

  • The extension adds vehicle-obstacle modeling to VANET simulators that already provide detailed instantaneous mobility information.The missing capability is modeling the intensity at which vehicles obstruct LOS.
  • Vehicles and buildings are represented as rectangles, and a line is drawn between the TX and RX antenna positions.
  • An unobstructed line indicates LOS, whereas intersection with another rectangle indicates obstruction by a vehicle or building.Geographical information distinguishes the two obstruction types.
  • After the propagation condition is identified, the simulator uses the relevant model to calculate power loss.Fresnel-zone clearance, obstacle height, obstacle location, TX–RX distance, and wavelength can refine LOS/OLOS characterization.

IV. NETWORK SIMULATIONS

Network simulations compare the LOS/OLOS model with Cheng’s Nakagami model across highway distances and vehicle densities. Although their average received powers coincide, packet reception and inter-arrival behavior differ with distance and density.

  • The simulations use a 10 km, four-lane highway with Poisson vehicle arrivals producing densities of approximately 100, 60, and 40 vehicles/km.The corresponding mean inter-arrival times are 1 s, 2 s, and 3 s.
  • As distance increases, the probability of OLOS rises, and averaged power is computed by weighting LOS and OLOS received powers by their distance-dependent probabilities.
  • The LOS/OLOS model’s averaged received power coincides with Cheng’s Nakagami model.
  • Within 100 m, both channel models have equal packet reception performance; at 200–400 m, Nakagami provides better packet reception probability.
  • For 40 vehicles/km, Nakagami gives better packet inter-arrival behavior at 200–300 m, while LOS/OLOS is slightly better above 400 m.

V. SUMMARY AND CONCLUSIONS

The paper presents a measurement-based shadow fading model that separates LOS, vehicle-obstructed LOS (OLOS), and building-obstructed LOS (NLOS) conditions. Vehicle obstruction adds about 10 dB of received-power loss, and simulations show performance differences from conventional Nakagami-based models.

  • The model separates measurements into LOS, vehicle-obstructed LOS (OLOS), and building-obstructed LOS (NLOS) conditions.
  • 10 dB is the approximate additional received-power loss induced by vehicles obstructing the LOS.
  • The LOS/OLOS and conventional Nakagami-based channel models produce different VANET simulation performance.
  • The LOS/OLOS model uses a dual-piecewise path-loss model and log-normal correlated shadowing whose mean depends on LOS or OLOS.
Loading 1203.3370v5…