Fuzzy-Logic Based Distributed Energy-Efficient Clustering Algorithm for Wireless Sensor Networks

Ying Zhang, Ying Zhang 0005, Jun Wang, Dezhi Han, Huafeng Wu, Rundong ZhouSensors 2017 · DOI

TL;DR

This paper introduces EEDCF, a novel fuzzy-logic-based distributed clustering algorithm for wireless sensor networks (WSNs). By incorporating residual energy, node degree, and crucially, neighbor node residual energy into a TSK fuzzy inference system, EEDCF significantly extends network lifetime and improves data transmission efficiency compared to existing methods.

What Problem Does This Paper Address?

Wireless Sensor Networks (WSNs) are increasingly deployed for various monitoring tasks, but their limited battery power poses a significant challenge. Clustering is a common strategy to manage energy consumption by electing Cluster Heads (CHs) to aggregate data from member nodes. However, traditional clustering algorithms often suffer from uneven energy distribution, particularly in multi-hop communication within clusters. Nodes closer to the CH act as relays and are drained faster, leading to premature network failure and impacting load balancing. Existing solutions often overlook crucial factors like node degree or rely on centralized approaches, which are not scalable for large WSNs. This paper aims to address these limitations by proposing a distributed, energy-efficient clustering algorithm that optimizes CH election by considering a more comprehensive set of node parameters.

Methods

The core of the proposed EEDCF algorithm lies in its distributed CH election mechanism, which leverages a Takagi-Sugeno-Kang (TSK) fuzzy inference system. Instead of traditional methods, EEDCF uses a TSK model for its quantitative analysis and efficient computation. The algorithm considers three key input parameters for each node:

  • Node Residual Energy (NE): Higher residual energy increases the probability of being elected as CH.
  • Node Degree (ND): A higher degree (more neighbors within communication radius) indicates better data transmission efficiency and reduced workload for individual relay nodes.
  • Neighbor Nodes' Average Residual Energy (NNE): This is a novel addition that addresses the load balancing issue in multi-hop communication. By considering the energy status of neighboring nodes, EEDCF aims to distribute the relaying burden more evenly, preventing hotspots and prolonging the network's overall lifetime.

The fuzzy inference system converts these crisp inputs into fuzzy linguistic variables (e.g., "low," "medium," "high" for energy; "less," "average," "enormous" for degree; "weak," "normal," "strong" for neighbor energy). These variables are then used in a set of IF-THEN rules, derived from empirical data and expert knowledge, to calculate a probability score for each node to become a CH. The node with the highest probability is elected as the CH. The algorithm operates in three phases: information table updating, CH competition, and cluster building.

Datasets

Not discussed in the paper. The authors conducted simulations using MATLAB.

Key Insights and Ideas

  • The Importance of Neighbor Node Energy: The most significant insight is the introduction of neighbor nodes' average residual energy as a crucial factor in CH election. This directly tackles the hotspot problem in multi-hop communication, which is often overlooked, and significantly improves load balancing.
  • TSK Fuzzy Model for Quantitative Analysis: Employing the TSK fuzzy model over traditional Mamdani models allows for a more direct quantitative output (probability) and simplifies the defuzzification process, making the algorithm more efficient.
  • Distributed Nature is Key: The algorithm's distributed nature means each node makes its CH election decision based on local information, reducing communication overhead and making it highly scalable for large WSNs.
  • Balancing Multiple Factors: EEDCF demonstrates that a balanced consideration of node energy, node degree, and neighbor node energy leads to superior performance compared to algorithms that focus on only one or two of these factors.

Key Results

The paper presents simulation results comparing EEDCF with two other prominent algorithms, EADEEG and DFLC, in two scenarios with varying node densities.

  • Network Lifetime: EEDCF consistently outperforms EADEEG and DFLC in extending network lifetime. In Scenario 1, EEDCF achieved an All Nodes Die (AND) lifetime of 3783 rounds, which is 11.6% longer than EADEEG and 7.7% longer than DFLC. In Scenario 2 (higher node density), EEDCF's AND lifetime was 3712 rounds, showing a significant improvement of 21.8% over EADEEG and 18.5% over DFLC.
  • Data Transmission Efficiency: EEDCF also shows superior data transmission efficiency. In Scenario 1, it transmitted 14.7 × 10⁴ Bytes, a 44.3% improvement over DFLC and 123.2% over EADEEG. In Scenario 2, this increased to 22.9 × 10⁴ Bytes, an 80.3% improvement over DFLC and 193.6% over EADEEG.
  • Energy Consumption: Figures 11 and 12 clearly show that EEDCF has the lowest system energy consumption throughout the network's life cycle, with residual energy gaps widening over time compared to the other algorithms.

Limitations and Future Work

  • Distance to Base Station: The paper acknowledges that EEDCF does not consider the distance to the Base Station (BS) as an input parameter for CH election. While this saves energy by avoiding real-time distance calculations, it might lead to CHs being farther from the BS. The authors argue that for distributed clustering, this is acceptable and can even economize energy. However, they suggest that for large-scale networks, the impact of this choice on overall energy consumption needs further investigation.
  • Security: The authors mention that future work will focus on improving the algorithm for security routing problems, aiming to establish trust evaluation models to defend against malicious nodes.
  • End-to-End Delay: Future research will also explore building secure links between nodes and CHs to effectively reduce end-to-end delay and improve overall network performance.

Summary updated

BibTeX
@article{zhang2017fuzzy,
  author = {Ying Zhang and Ying Zhang and Jun Wang and Dezhi Han and Huafeng Wu and Rundong Zhou},
  title = {{Fuzzy-Logic Based Distributed Energy-Efficient Clustering Algorithm for Wireless Sensor Networks}},
  journal = {Sensors},
  year = {2017},
  doi = {10.3390/s17071554},
}
Figure