Multi-Attribute Fusion Algorithm Based on Improved Evidence Theory and Clustering

Wenqing Wang, Yuan Yan, Rundong Zhang, Zhen Wang, Yongqing Fan, Chunjie Yang

TL;DR

This paper introduces a novel multi-attribute fusion algorithm that combines fuzzy clustering with an improved Dempster-Shafer (DS) evidence theory to dynamically adapt the switching threshold for smart street lighting systems. By intelligently fusing sensor data on ambient and street light illumination, the algorithm achieves more accurate and energy-efficient lighting control compared to fixed-threshold methods.

What Problem Does This Paper Address?

Traditional industrial control systems, like street lighting, often rely on fixed switching thresholds. This approach is inefficient and inflexible, especially in the context of smart cities that demand adaptive and optimized resource management. The challenge lies in accurately determining the optimal moment to switch lights on or off, considering multiple environmental factors that influence illumination levels. Existing methods often assume sensor data follows normal distributions with consistent parameters, which may not hold true due to real-world sensor errors, environmental variations, and other interferences. This paper aims to address this by developing a fusion algorithm that can handle diverse sensor observations and derive a more precise, adaptive switching threshold.

Methods

The proposed algorithm integrates two key techniques:

  • Fuzzy C-Means (FCM) Clustering: This algorithm is used to group sensor observations into distinct categories. Unlike traditional K-means, FCM assigns a membership grade to each data point for every cluster, allowing for more nuanced data partitioning. The authors selected FCM because it offers higher accuracy than classical K-means and can automatically group observations without requiring prior knowledge of category numbers (though the number of clusters is specified in advance for the experiment).
  • Improved Dempster-Shafer (DS) Evidence Theory: After clustering, the DS evidence theory is applied to fuse the data within each cluster. This theory is well-suited for handling uncertainty and incomplete information. The authors introduce an "improved" version that incorporates a reliability factor and a conflict factor to better manage conflicting evidence and improve fusion accuracy. The process involves converting observations into evidence, judging their accuracy, and then combining this evidence using a fusion rule.

Datasets

The experiments were conducted using a dataset of 100 two-dimensional observations collected from a smart campus lighting system. These observations comprised:

  • Ambient Illumination: Measured in Lux.
  • Street Light Illumination: Measured in Lux.

These datasets were collected within a 5-minute window around dusk, capturing the rapid changes in light levels. The paper does not introduce a new benchmark dataset but uses real-world data from a specific application scenario.

Key Insights and Ideas

  • Adaptive Thresholding is Crucial for Smart Cities: The paper strongly advocates for adaptive thresholds over fixed ones, highlighting the potential for energy savings and improved efficiency in smart city applications.
  • FCM for Robust Data Grouping: Using FCM for initial data clustering proves effective in handling the variability and potential overlap in sensor readings, creating more meaningful groups for subsequent fusion.
  • DS Evidence Theory for Handling Uncertainty: The DS evidence theory, particularly with the proposed improvements, offers a robust framework for combining uncertain and potentially conflicting sensor data, leading to more reliable fusion results.
  • The Importance of Cluster Number Selection: The study demonstrates that the choice of the number of clusters (two, three, or four) significantly impacts the final fusion result, suggesting that an optimal number of clusters is critical for achieving the best performance.

Key Results

The study evaluated the proposed algorithm by clustering the 100 two-dimensional datasets into two, three, and four categories. The fusion results from each clustering case were then compared.

  • Optimal Clustering: The analysis indicated that clustering the data into three categories yielded the best fusion result with a relative error of 2.003%.
  • Comparison with Fixed Threshold: The proposed adaptive fusion method achieved a fusion result of 20.4006 (with a 2.003% error) when using three clusters. This is contrasted with a fixed threshold method, which, when applied to the same data, produced significantly different and less accurate results (e.g., 34.8223 for one group with two clusters).
  • High Accuracy: The paper claims that the correct rate of the proposed algorithm is above 97.997%, outperforming a similar method from prior research.

Limitations and Future Work

The authors acknowledge several limitations and suggest future research directions:

  • Clustering Method Limitations: The FCM algorithm can fall into local optimal solutions, especially with large datasets. It also requires pre-defining the number of cluster centers.
  • Distinguishing Signal from Noise: A key challenge is effectively distinguishing useful information from noise during the sample classification process.
  • Alternative Clustering Algorithms: The paper suggests exploring other algorithms, such as Bayesian estimation, for more accurate initial observation clustering.
  • Intelligent Cluster Center Adaptation: Future work could focus on developing methods for the algorithm to intelligently adapt its cluster centers for different application problems.
  • Extension to Multidimensional Attributes: The current work focuses on two-dimensional attributes, but the authors suggest extending the algorithm to handle multidimensional attributes for broader applications like traffic control, battlefield situation estimation, and target tracking.

Summary updated

Abstract

In most of the application scenarios of industrial control systems, the switching threshold of a device, such as a street light system, is typically set to a fixed value. To meet the requirements for a smart city, it is necessary to set a threshold that is adaptive to different conditions by fusing the multi-attribute observations of the sensors. This paper proposes a multi-attribute fusion algorithm based on fuzzy clustering and improved evidence theory. All of the observations are clustered by fuzzy clustering, where a proper clustering method is chosen, and the improved evidence theory is used to fuse the observations. In the experiments, two-dimensional observations for the street light illumination and for the ambient illumination are used in a campus-intelligent lighting system based on a narrowband Internet of things, and the results demonstrate the effectiveness of the proposed fusion algorithm. The proposed algorithm can be applied to a variety of multi-attribute fusion scenarios.
BibTeX
@inproceedings{wang2019multi,
  author = {Wenqing Wang and Yuan Yan and Rundong Zhang and Zhen Wang and Yongqing Fan and Chunjie Yang},
  title = {{Multi-Attribute Fusion Algorithm Based on Improved Evidence Theory and Clustering}},
  booktitle = {Sensors},
  year = {2019},
  url = {https://openreview.net/forum?id=5GaF4xMnhw},
}
Figure