Correlated Spatio-Temporal Data Collection in Wireless Sensor Networks Based on Low Rank Matrix Approximation and Optimized Node Sampling

Xinglin Piao, Yongli Hu, Yanfeng Sun, Baocai Yin, Junbin Gao

TL;DR

This paper introduces a novel method for efficient data collection in wireless sensor networks (WSNs) by leveraging both spatial and temporal correlations in sensor data using a Correlated Spatio-Temporal Low Rank Matrix Approximation (ST-LRMA) model. It also proposes an optimized node sampling strategy using the Gini index to balance data recovery accuracy with network energy consumption, thereby extending network lifetime.

What Problem Does This Paper Address?

Wireless Sensor Networks (WSNs) are crucial for environmental monitoring and other applications, but sensor nodes have limited power. Efficient data collection is paramount to prolonging network lifetime. While Low Rank Matrix Approximation (LRMA) offers a promising approach by recovering data from a subset of sensed information, existing LRMA methods often fail to fully exploit the inherent spatial and temporal correlations in sensor data. This leads to suboptimal data recovery and, more critically, uneven energy distribution among nodes due to the many-to-one data collection pattern, ultimately shortening the network's operational lifespan. The paper aims to address these limitations by developing a method that simultaneously considers data correlations and optimizes node sampling for energy efficiency.

Methods

The core of the proposed solution lies in two main components:

  • Correlated Spatio-Temporal Low Rank Matrix Approximation (ST-LRMA): This novel LRMA model integrates both temporal consistency (data from a single sensor over time) and spatial correlation (data from neighboring sensors at the same time) into the matrix completion process. This is achieved by augmenting the standard LRMA objective function with terms that penalize deviations from temporal smoothness (using matrix D) and spatial similarity (using matrix S). The matrix S is constructed using a Sparse Manifold Clustering (SMC) method to capture complex spatial relationships, even when node locations are not precisely known. The ST-LRMA model is then solved using an iterative optimization algorithm.
  • Optimized Node Sampling: To address the energy imbalance caused by traditional random sampling, the paper proposes an optimized node sampling strategy. It uses the Gini index to quantify two key aspects:
  • Spatial Distribution: The uniformity of selected sensor nodes across the network.
  • Network Energy Status: The evenness of remaining energy among nodes. These two Gini indices are then combined into a multi-objective optimization problem, solved using Particle Swarm Optimization (PSO), to select a subset of nodes for sensing that balances data recovery quality with energy distribution, thus maximizing network lifetime.

Datasets

  • Simulated Network: A 21x21 grid network was simulated to generate ground truth data. This data was then augmented with Gaussian noise to mimic real-world sensing conditions.
  • Real-World Network: Data from the Intel Berkeley Research Lab wireless sensor network was used. This dataset includes measurements of temperature, humidity, and illumination from 53 nodes over several months.

Key Insights and Ideas

  • Synergy of Spatio-Temporal Correlation: The paper demonstrates that jointly considering spatial and temporal correlations in LRMA leads to significantly better data recovery accuracy compared to methods that only consider one or the other.
  • Gini Index for Energy-Aware Sampling: The Gini index proves to be an effective tool for quantifying both spatial node distribution and energy evenness, enabling a more intelligent and energy-efficient node selection process than simple random sampling.
  • Optimized Sampling Enhances Lifetime: By balancing node selection for data quality and energy distribution, the proposed sampling method can significantly prolong network lifetime, especially at lower sampling rates.
  • Robustness to Data Loss: The ST-LRMA method shows remarkable stability and accuracy even when entire rows and columns of data are missing, a common scenario in WSNs.

Key Results

  • Improved Data Recovery Accuracy: The proposed ST-LRMA method consistently outperformed conventional interpolation methods and other LRMA-based approaches (B-LRMA, T-LRMA, S-LRMA) in terms of data recovery accuracy, especially at low sampling rates. For instance, in simulated networks, ST-LRMA achieved significantly lower relative errors.
  • Extended Network Lifetime: The optimized node sampling strategy, when integrated with ST-LRMA, led to a substantial increase in network lifetime. In simulations, for sampling rates below 30%, the optimized method extended the network lifetime by up to two times compared to random sampling.
  • Robustness: The ST-LRMA method demonstrated stable recovery results even when 10% of rows and columns were randomly removed from the measurement matrices, while other methods showed large errors and un-convergent behavior.

Limitations and Future Work

  • Network Properties Not Fully Considered: The current experiments did not account for complex routing protocols, network time duty cycles, or specific sensing data attributes.
  • Future Work: The authors suggest integrating these network properties into the ST-LRMA method for application on larger-scale WSNs. Additionally, they propose exploring multi-modality sensing data and formulating sensing data as tensors for recovery using low-rank tensor constraints.

Summary updated

Abstract

The emerging low rank matrix approximation (LRMA) method provides an energy efficient scheme for data collection in wireless sensor networks (WSNs) by randomly sampling a subset of sensor nodes for data sensing. However, the existing LRMA based methods generally underutilize the spatial or temporal correlation of the sensing data, resulting in uneven energy consumption and thus shortening the network lifetime. In this paper, we propose a correlated spatio-temporal data collection method for WSNs based on LRMA. In the proposed method, both the temporal consistence and the spatial correlation of the sensing data are simultaneously integrated under a new LRMA model. Moreover, the network energy consumption issue is considered in the node sampling procedure. We use Gini index to measure both the spatial distribution of the selected nodes and the evenness of the network energy status, then formulate and resolve an optimization problem to achieve optimized node sampling. The proposed method is evaluated on both the simulated and real wireless networks and compared with state-of-the-art methods. The experimental results show the proposed method efficiently reduces the energy consumption of network and prolongs the network lifetime with high data recovery accuracy and good stability.
BibTeX
@inproceedings{piao2014correlated,
  author = {Xinglin Piao and Yongli Hu and Yanfeng Sun and Baocai Yin and Junbin Gao},
  title = {{Correlated Spatio-Temporal Data Collection in Wireless Sensor Networks Based on Low Rank Matrix Approximation and Optimized Node Sampling}},
  booktitle = {Sensors},
  year = {2014},
  url = {https://openreview.net/forum?id=1RY6TCSgn-O},
}
Figure