Measurement and Analysis of Near-Ground Propagation Models under Different Terrains for Wireless Sensor Networks

Weisheng Tang, Xiaoyuan Ma, Jianming Wei, Zhi WangSensors 2019 · DOI

TL;DR

This paper addresses the critical need for accurate near-ground propagation models in Wireless Sensor Networks (WSNs), which often have antennas just centimeters off the ground. By conducting real-world measurements across various terrains and antenna heights, the authors developed new statistical models and a selection strategy, demonstrating that simplistic models lead to significant inaccuracies in predicting network performance.

What Problem Does This Paper Address?

Wireless Sensor Networks (WSNs) are increasingly deployed for diverse applications, from infrastructure monitoring to environmental sensing. A fundamental aspect of designing and deploying these networks is understanding how radio signals propagate. This is typically done using propagation models, which predict signal strength and thus network coverage and connectivity. However, most existing propagation models are based on assumptions that don't hold for WSNs, particularly when sensor nodes are placed directly on the ground with antennas at very low heights (a few centimeters). This "near-ground" scenario, common in WSN deployments, is poorly represented by traditional models, leading to inaccurate predictions of network performance and potentially unreliable deployments. The paper aims to bridge this gap by providing accurate near-ground propagation models and a method for selecting the most appropriate one for a given scenario.

Methods

The core of this research involves empirical measurements and subsequent model development. The authors designed and built custom WSN nodes using Silicon Labs Si4432 RF chips and MSP430F5438 MCUs, operating at 470 MHz with a transmission power of +20 dBm and 0 dBi antenna gain.

Measurement Campaign: * Terrains: Four distinct scenarios were used: flat concrete road, flat grass, undulating grass, and flat grass with obstacles. * Antenna Heights: Three antenna heights were tested: 5 cm (representing nodes directly on the ground), 50 cm, and 1 m. * Data Collection: Received Signal Strength Indication (RSSI) was measured. The transmitter was fixed, and the receiver was moved away in steps of 1m up to 10m, then 2m up to 20m, and finally 5m up to 50m. At each point, 200 packets were collected to ensure statistical robustness. * Path Loss Calculation: RSSI values were converted to received power, and path loss was calculated using the known transmitted power.

Modeling: * Statistical Models: Based on the observed log-normal attenuation patterns, both one-slope and two-slope log-normal shadowing models were fitted to the measured path loss data using least squares fitting. The breakpoint for the two-slope model was calculated using a derived formula. * Model Comparison: The accuracy of the derived statistical models was compared against standard theoretical models like Free Space Path Loss (FSPL), two-ray (plane earth), and empirical models like COST231-Walfish-Ikegami. * Simulation: The developed propagation models were implemented in the Cooja simulator to evaluate their impact on network connectivity.

Propagation Model Selection Strategy: * A flowchart-based strategy was proposed to guide users in selecting the most appropriate propagation model based on whether actual measurements are feasible, the availability of empirical models for similar conditions, and the antenna height.

Datasets

The paper does not introduce new benchmark datasets in the traditional sense. Instead, it presents a newly collected dataset of near-ground radio propagation measurements. This dataset comprises RSSI measurements taken at 470 MHz across four different terrains and three antenna heights, covering distances up to 50 meters. This empirical dataset serves as the basis for deriving and validating the proposed propagation models.

Key Insights and Ideas

  • Antenna Height is Crucial: The most significant finding is the profound impact of antenna height on near-ground propagation. Even small changes, from 5 cm to 1 m, drastically alter path loss and thus network coverage. This highlights why standard models fail for WSNs.
  • Near-Ground is Not Free Space: The FSPL model is highly optimistic and completely unsuitable for near-ground WSN scenarios, predicting much lower path loss than observed.
  • Two-Ray Model's Limitations: While the two-ray model is more realistic than FSPL by considering ground reflection, it becomes inaccurate at very low antenna heights and is highly sensitive to height changes, making it unreliable for many WSN deployments.
  • Statistical Models Excel: Log-normal shadowing models, particularly the two-slope variant for higher antenna heights, provide a much better fit to the measured near-ground propagation characteristics.
  • Terrain Matters: Different terrains (concrete vs. grass, with or without obstacles) exhibit distinct path loss behaviors, necessitating terrain-specific or adaptable models.
  • Simplistic Models Distort Network Performance: Simulating WSNs with inaccurate propagation models (like FSPL or two-ray) leads to severely overestimated network connectivity, masking real-world limitations.

Key Results

  • Developed Statistical Models: The paper successfully derived one-slope and two-slope log-normal shadowing models for near-ground propagation under various terrains and antenna heights.
  • Model Performance: The two-slope model generally offered higher precision (lower RMSE and MAPE) than the one-slope model, especially as antenna height increased, except for extremely low heights (below 50 cm) where the one-slope model was sufficient and simpler.
  • Comparison with Theoretical Models: FSPL and COST231-Walfish-Ikegami models significantly underestimated path loss. The two-ray model showed varying accuracy, being closer at 5cm but becoming inaccurate at 1m.
  • Simulation Impact: When implemented in Cooja, simplistic models like FSPL predicted near-perfect connectivity (100%), while the measured models showed significantly lower connectivity, especially at 5cm antenna height (around 15% for some scenarios). This demonstrates the critical need for accurate models in simulation.

Limitations and Future Work

Limitations: * Frequency Specificity: The measurements and derived models are specific to the 470 MHz frequency band. Extrapolating these models to other frequencies might require further validation. * Limited Terrains: While four scenarios were tested, the diversity of real-world environments is vast. The models might need refinement for other complex terrains (e.g., dense urban, forest canopy). * Measurement Range: The measurements extended up to 50 meters. For longer-range WSN deployments, additional data would be beneficial. * Antenna Characteristics: While antenna gain was considered, detailed antenna radiation patterns were not explicitly modeled, which could influence propagation in specific orientations.

Future Work: * Broader Frequency and Terrain Coverage: Expanding measurements to cover a wider range of frequencies and more diverse terrains is suggested. * Dynamic Environments: Investigating the impact of dynamic environmental factors (e.g., moving objects, weather changes) on near-ground propagation. * Advanced Modeling: Exploring more sophisticated models that can adapt to varying terrain and environmental conditions, potentially incorporating machine learning techniques. * Protocol-Level Impact: Further in-depth analysis of how these accurate models affect the performance of specific WSN routing and MAC protocols.

Summary updated

Abstract

The propagation model is an essential component in the design and deployment of a wireless sensor network (WSN). Although much attention has been given to near-ground propagation models, few studies place the transceiver directly on the ground with the height of antennas at the level of a few centimeters, which is a more realistic deployment scenario for WSNs. We measured the Received Signal Strength Indication (RSSI) of these truly near-ground WSNs at 470 MHz under four different terrains, namely flat concrete road, flat grass and two derived scenarios, and obtained the corresponding path loss models. By comprehensive analysis of the influence of different antenna heights and terrain factors, we showed the limit of existing theoretical models and proposed a propagation model selection strategy to more accurately reflect the true characteristics of the near-ground wireless channels for WSNs. In addition, we implemented these models on Cooja simulator and showed that simplistic theoretical models would induce great inaccuracy of network connectivity estimation.
BibTeX
@inproceedings{tang2019measurement,
  author = {Weisheng Tang and Xiaoyuan Ma and Jianming Wei and Zhi Wang},
  title = {{Measurement and Analysis of Near-Ground Propagation Models under Different Terrains for Wireless Sensor Networks}},
  booktitle = {Sensors},
  year = {2019},
  doi = {10.3390/s19081901},
}
Figure