Enhance the Quality of Crowdsensing for Fine-Grained Urban Environment Monitoring via Data Correlation

Xu Kang, Liang Liu, Huadong Ma

TL;DR

This paper tackles the challenge of low-quality urban environment sensing data due to sparse and unevenly distributed mobile sensors. They propose a novel crowdsensing approach that leverages both temporal and cross-sensor data correlations using a collaborative tensor decomposition method to significantly improve the accuracy and detail of urban environment monitoring.

What Problem Does This Paper Address?

Urban environments are complex and dynamic, requiring detailed monitoring for effective management. Traditional methods like Urban Remote Sensing (URS) lack fine-grained precision, while Wireless Sensor Networks (WSNs) suffer from energy and stability issues. "Crowdsensing," utilizing ubiquitous mobile devices like smartphones and vehicles, offers a promising alternative for city-scale monitoring. However, crowdsensing data is inherently sparse and unevenly distributed, leading to "blank zones" and reduced sensing quality. This paper aims to enhance the quality of these sensing images, effectively increasing the "urban resolution" of the crowdsensing network, especially in these data-scarce areas.

Methods

The authors propose a "data correlation based crowdsensing approach" to improve sensing image quality. Their method involves:

  • Data Modeling: Representing sensory data as a 3D tensor with dimensions for region, signal category (target and correlated signals), and time slots.
  • Signal Correlation Matrix: Quantifying the correlation between different environmental phenomena (e.g., PM2.5 and PM10).
  • Collaborative Tensor Decomposition: A novel approach that simultaneously decomposes the sparse sensory data tensor and a signal correlation matrix. This allows the model to "borrow" information from correlated signals to fill in missing data in the target signal, especially in blank zones. They use a loss function that balances reconstruction error with regularization.
  • Correlated Time Slots Combination: Aggregating data from multiple continuous time slots to further enhance sensing in regions where data is sparse over time. This leverages the temporal stability of environmental phenomena.

The core innovation lies in the collaborative tensor decomposition, which jointly models sensor data and inter-signal correlations, and the combination of temporal and cross-sensor data fusion.

Datasets

  • Simulated Data: The authors generated synthetic mobility traces using the Self-similar Least Action Walk (SLAW) model [11] to simulate the movement of 6000 crowdsensing nodes. They also generated synthetic target (CO2 concentration) and correlated environmental signals (linear and non-linear functions of the target signal with added noise).
  • Real-World Data: Air quality monitoring data from Beijing, specifically PM2.5, PM10, and air humidity index from 22 monitoring stations, collected between April 21-27, 2013. This data was used for a case study to evaluate the approach in a real-world scenario.

Key Insights and Ideas

  • The "Urban Camera" Analogy: Crowdsensing networks can be viewed as an "urban camera," where the quality of the captured "image" (urban status) is directly related to the density and distribution of sensors.
  • Data Sparsity is the Bottleneck: The primary challenge in crowdsensing is data sparsity, caused by uneven participant distribution, leading to "blank zones" with poor sensing.
  • Correlation is Key: Both temporal (how a phenomenon changes over time) and category correlation (how different phenomena relate to each other) are crucial for filling in missing data and enhancing sensing quality.
  • Tensor Decomposition for Multi-dimensional Data: Tensors are a natural fit for modeling spatio-temporal and multi-category environmental data, and tensor decomposition provides a powerful tool for inferring missing values.
  • Collaborative Learning Across Signals: The most significant insight is that by jointly modeling the target signal and its correlated counterparts, the system can achieve better reconstruction than by treating each signal in isolation. This is akin to using multiple related "lenses" to get a clearer picture.

Key Results

  • Improved Sensing Quality: The proposed enhanced crowdsensing approach significantly improves the quality of generated sensing images compared to traditional spatial interpolation methods.
  • Higher "Urban Resolution": The paper introduces "crowdsensing resolution" as a metric for sensing ability. Their method achieves higher resolution (quantified by the parameter 'a' in the linear relationship between resolution and the number of nodes) than the interpolation method. For example, in simulations, the resolution parameter 'a' increased from 0.583 for interpolation to 0.774 and 0.831 for their enhanced approaches with different time slot combinations.
  • Validation with Real Data: In the Beijing air quality case study, the enhanced approach generated a sensing image with more details and a higher correlation coefficient (0.91) with the ground truth compared to the interpolation method (0.58).

Limitations and Future Work

  • Limitations:
    • The performance of the collaborative tensor decomposition relies on the existence and strength of correlations between different environmental signals. If signals are uncorrelated, the benefit will be limited.
  • The method assumes that the environmental phenomenon remains relatively stable within a time slot.
  • The computational complexity of tensor decomposition can be high, although the paper uses techniques to manage this.
  • Future Work:
    • The authors suggest exploring more sophisticated mobility models and environmental signal generation models.
  • Investigating the impact of different sensor types and their specific error characteristics.
  • Further research into optimizing the tensor decomposition process for even greater efficiency.
  • Extending the approach to handle more complex urban environments and a wider range of environmental phenomena.

Summary updated

Abstract

Monitoring the status of urban environments, which provides fundamental information for a city, yields crucial insights into various fields of urban research. Recently, with the popularity of smartphones and vehicles equipped with onboard sensors, a people-centric scheme, namely “crowdsensing”, for city-scale environment monitoring is emerging. This paper proposes a data correlation based crowdsensing approach for fine-grained urban environment monitoring. To demonstrate urban status, we generate sensing images via crowdsensing network, and then enhance the quality of sensing images via data correlation. Specifically, to achieve a higher quality of sensing images, we not only utilize temporal correlation of mobile sensing nodes but also fuse the sensory data with correlated environment data by introducing a collective tensor decomposition approach. Finally, we conduct a series of numerical simulations and a real dataset based case study. The results validate that our approach outperforms the traditional spatial interpolation-based method.
BibTeX
@inproceedings{kang2017enhance,
  author = {Xu Kang and Liang Liu and Huadong Ma},
  title = {{Enhance the Quality of Crowdsensing for Fine-Grained Urban Environment Monitoring via Data Correlation}},
  booktitle = {Sensors},
  year = {2017},
  url = {https://openreview.net/forum?id=HfrAw4JQT9},
}
Figure