An Efficient Information Maximization Based Adaptive Congestion Control Scheme in Wireless Sensor Network

Jiawei Tan, Wei Liu, Tian Wang, Shaobo Zhang, Anfeng Liu, Mande Xie, Ming Ma, Ming Zhao

Introduction

Wireless Sensor Networks (WSNs) serve as the foundational infrastructure for the Internet of Things (IoT), enabling real-time monitoring across diverse domains such as industrial automation and environmental conservation. However, WSNs are inherently resource-constrained. As node density and data traffic increase, networks frequently encounter "bottlenecks," leading to severe congestion. Traditional congestion control mechanisms often fall short, either by dropping critical data packets indiscriminately or by routing data through longer paths that deplete energy and introduce significant latency.

The Challenge of Content-Agnostic Control

Existing solutions typically employ rate control or detour routing. Rate control often fails to distinguish between redundant "steady-state" data and critical "event" alerts, leading to the loss of high-value information. Conversely, detour routing increases path length, which elevates energy consumption and transmission delays. The fundamental issue is that these methods manage raw traffic volume rather than the actual information content being transmitted.

The DRCDC Scheme: Prioritizing Information

To address these limitations, researchers have proposed the Differentiated Rate Control Data Collection (DRCDC) scheme. This approach shifts the focus from managing packet quantity to maximizing the quality of information (QoI) collected at the sink. DRCDC utilizes two primary mathematical frameworks to mitigate congestion:

1. Spatial Mitigation via Matrix Completion

WSN data typically exhibits high spatial correlation; sensors in close proximity often record similar values. DRCDC leverages Matrix Completion Theory to exploit this redundancy. By treating the sensor data as a low-rank matrix, the sink can mathematically reconstruct missing data points (the "holes" in the matrix). When a node experiences congestion, it can intentionally reduce its transmission rate, knowing that the sink can "fill in the blanks" using data from neighboring, non-congested nodes.

2. Temporal Mitigation via Information Entropy

The scheme also calculates the "temporal entropy" of sensor data, measuring the novelty of each reading relative to previous ones. If a sensor's readings remain stable over time, the resulting packets have low entropy and are prioritized for rate reduction or dropping during congestion. This ensures that only the most informative data—representing significant changes or events—is prioritized for transmission.

Adaptive Rate Adjustment

DRCDC employs specific algorithms to detect congestion through queue thresholds and adaptively adjust transmission rates. A unique feature of this scheme is its compensatory mechanism: when a congested node reduces its rate, the algorithm allows non-congested nodes to slightly increase their rates. This ensures that the overall data matrix remains "recoverable" at the sink, maintaining high data fidelity despite local network stress.

Experimental Validation

The performance of DRCDC was evaluated using two distinct datasets:

  • Intel Berkeley Research Lab Dataset: Real-world traces from 52 nodes were used to test the accuracy of matrix-based data reconstruction.
  • OMNeT++ Simulation: A large-scale simulation involving 500 nodes was conducted to measure network-wide metrics such as delay and packet loss.

Key Results and Performance

When compared against traditional No Congestion Control (NCCS) and Random Uniform Distribution (RUDS) schemes, DRCDC demonstrated significant improvements:

  • Massive Delay Reduction: Maximum network delay was slashed by 34.2% to 76.3%.
  • Improved Reliability: Packet loss rates were reduced by 2.8% to 6.8%.
  • Optimized Information Density: The average entropy of collected data was lower, indicating a higher concentration of useful information per packet.
  • Enhanced Network Longevity: By alleviating pressure on bottleneck nodes, the scheme prevents early node failure, thereby extending the overall operational lifetime of the network.

Conclusion and Future Directions

The DRCDC scheme represents a paradigm shift in WSN management by treating data as information rather than just bits. While highly effective in correlated environments, future work will focus on optimizing "spatial entropy" control and testing the scheme's robustness in highly dynamic or mobile network topologies where the "low-rank" assumption of data may be challenged.

Summary updated

Abstract

Intelligent communication technology can enhance user experience and improve people's lifestyle. However, due to the limited resources of wireless sensor networks (WSNs), the congestion occurs when the node's traffic load exceeds it's the available capacity. Congestion may cause serious problems such as high packet loss and low throughput, which is an extremely deleterious impact on the performance of WSNs. To tackle this issue, a differentiated rate control data collection (DRCDC) scheme is proposed to avoid or alleviate the network congestion. Different from the previous scheme, the DRCDC scheme can mitigate the congestion by intelligently reducing the data that contains less information, and maintain the distortion rate of information collected by the network is small. The DRCDC mainly avoids or mitigates congestion on the spatial and temporal through information entropy theory, and at the same time makes the information collected by the network has the lowest distortion rate. 1) Congestion mitigation in spatial. Since there is a spatial correlation between sensing data, so in the case of collecting a certain amount of data, some missing data in space can be recovered from another spatially acquired data by matrix completion theory. Therefore, in the DRCDC scheme, when congestion occurs, the amount of data transmitted (forwarded) for nodes are congested will be reduced by an intelligent approach, thereby being able to mitigate congestion. 2) Congestion mitigation in temporal. when congestion occurs, reducing the data collection of time slots containing small information to reduce the flow rate that needs to be transmitted, thereby avoiding congestion while minimizing data distortion rate of the network. The experimental results of the DRCDC scheme in a planar network show better performance than the traditional data collection schemes and reducing the packet loss 2.8%-6.8% while reduce the maximum delay by 34.2%-76.3%.
BibTeX
@article{tan2019efficient,
  author = {Jiawei Tan and Wei Liu and Tian Wang and Shaobo Zhang and Anfeng Liu and Mande Xie and Ming Ma and Ming Zhao},
  title = {{An Efficient Information Maximization Based Adaptive Congestion Control Scheme in Wireless Sensor Network}},
  journal = {IEEE Access},
  year = {2019},
  doi = {10.1109/access.2019.2915385},
}
Figure