On the optimum number of cooperating nodes in interfered cluster-based sensor networks

Stefan Mijovic, Luca Sanguinetti, Chiara Buratti, Mérouane Debbah

TL;DR

This paper tackles the challenge of optimizing energy efficiency in wireless sensor networks (WSNs) by determining the ideal number of nodes per cluster. They propose a cooperative MIMO scheme that accounts for realistic hardware limitations and interference, finding that increasing interference and the number of sinks necessitates more nodes per cluster for optimal energy use.

What Problem Does This Paper Address?

Wireless sensor networks (WSNs) are increasingly deployed for various applications, but transmitting data over long distances directly from nodes to sinks is energy-intensive and inefficient. Cooperative MIMO (Multiple-Input Multiple-Output) offers a way to improve spectral efficiency by allowing nodes to act as a virtual antenna array. However, implementing this in WSNs faces hurdles like hardware constraints, imperfect synchronization, and interference between clusters. This paper specifically addresses how to find the "sweet spot" – the optimal number of sensor nodes within a cluster – to maximize the network's energy efficiency, considering these real-world challenges.

Methods

The authors propose a cooperative MIMO scheme for WSNs organized into clusters, where nodes within a cluster cooperate to transmit data to sinks. They employ a Wyner model to simplify the analysis of inter-cluster interference and focus on an asymptotic regime where the number of nodes per cluster (N) grows infinitely large. This allows for a closed-form solution for the optimal N. The analysis explicitly incorporates several realistic imperfections:

  • Hardware Impairments: Modeled as a reduction in signal strength and added Gaussian noise.
  • Imperfect Synchronization: Represented by random phase shifts, affecting signal quality and introducing interference.
  • Channel Estimation Errors: Acknowledged as a practical limitation.
  • Data Correlation: Nodes within a cluster transmit correlated data.
  • Inter-cluster Interference: Modeled based on the proximity of clusters.

They derive expressions for the achievable rate (SINR) and total power consumption, then optimize the energy efficiency (EE) – defined as the ratio of achievable rate to power consumption.

Datasets

Not discussed in the paper. The paper relies on theoretical analysis and simulations using parameters provided in Table I.

Key Insights and Ideas

  • The "Sweet Spot" for Cluster Size: There's an optimal number of nodes per cluster that balances the benefits of cooperation (improved SINR) against the increased power consumption and potential for interference.
  • Interference is a Driver for Cluster Size: Higher inter-cluster interference forces the system to use more nodes per cluster to maintain a good SINR, thus increasing the optimal N.
  • More Sinks Mean Bigger Clusters: An increase in the number of sinks per cluster also leads to a higher optimal N. This is because sink power consumption becomes a more significant part of the total energy budget, and increasing N helps improve the rate without drastically increasing this dominant power component.
  • Asymptotic Analysis is Surprisingly Accurate: The mathematical analysis performed in the limit of infinite nodes per cluster provides results that closely match simulations even for relatively small, finite numbers of nodes. This validates the approach and provides a practical tool for design.
  • Synchronization Matters: Imperfect synchronization significantly degrades energy efficiency, highlighting its importance in practical WSN deployments.

Key Results

  • The paper derives a closed-form expression for the energy efficiency (EE) as a function of the number of nodes per cluster (N), inter-cluster interference (α), and the number of sinks (K).
  • They derive a closed-form expression for the EE-optimal number of nodes per cluster, N*, given by $N^* = \frac{e^{(z^*+1)}-1}{\gamma_l}$, where $z^*$ is a function of the system parameters.
  • Simulations confirm that the asymptotic analysis accurately predicts the EE-optimal N for finite system sizes.
  • The optimal N* increases with both the inter-cluster interference parameter α and the number of sinks K.
  • The paper quantifies the performance degradation due to imperfect synchronization, showing a significant drop in EE as synchronization error increases.

Limitations and Future Work

  • Asymptotic Regime: While accurate for finite systems, the primary analysis is in the asymptotic regime (N → ∞). The paper acknowledges that the difference between asymptotic and finite results is higher for larger K.
  • Simplified Channel Model: The paper uses a block flat-fading channel model. More complex fading models could be considered.
  • Wyner Model Simplification: The Wyner model simplifies interference, and its applicability to more complex network topologies might be limited.
  • Future Work: The authors suggest that a more detailed analysis of imperfect synchronization's impact on performance is warranted, as this aspect is "yet to be analysed in details." They also note that their insights could be extended to other network parameters.

Summary updated

Abstract

This paper presents a cooperative multiple-input multiple-output (MIMO) scheme for a wireless sensor network consisting of inexpensive nodes, organised in clusters and transmitting data towards sinks. The transmission is affected by hardware imperfections, imperfect synchronisation, data correlation among nodes of the same cluster, channel estimation errors and interference among nodes of different clusters. Within this setting, we are interested in determining the number of nodes per cluster that maximises the energy efficiency of the network. The analysis is conducted in the asymptotic regime in which the number N of sensor nodes per cluster grows large without bound. Numerical results are used to validate the asymptotic analysis in the finite system regime and to investigate different configurations. It turns out that the optimum number of sensor nodes per cluster increases with the inter-cluster interference and with the number of sinks.
BibTeX
@inproceedings{mijovic2016optimum,
  author = {Stefan Mijovic and Luca Sanguinetti and Chiara Buratti and Mérouane Debbah},
  title = {{On the optimum number of cooperating nodes in interfered cluster-based sensor networks}},
  booktitle = {ICC},
  year = {2016},
  doi = {10.1109/icc.2016.7511418},
}
Figure