Towards Energy-Fairness in Asynchronous Duty-Cycling Sensor Networks

Zhenjiang Li, Mo Li, Yunhao LiuACM Trans. Sens. Networks 2014 · DOI

TL;DR

This paper tackles the critical challenge of ensuring energy fairness in asynchronous duty-cycling wireless sensor networks (WSNs). They propose a novel distributed algorithm, GDSIC, that dynamically adjusts node sleep intervals to minimize the maximum energy consumption rate across the network, thereby prolonging overall network lifetime.

What Problem Does This Paper Address?

Wireless Sensor Networks (WSNs) are increasingly deployed for long-term monitoring, but their battery-powered nature presents a significant challenge: limited energy. Duty-cycling, where nodes periodically sleep to conserve power, is a common solution. However, in asynchronous WSNs, where nodes don't strictly synchronize their sleep/wake cycles, controlling these sleep intervals to ensure energy fairness is difficult. Traditional approaches like setting identical sleep intervals or greedily adjusting them for nodes with high consumption rates fail to achieve optimal fairness and can lead to premature node failures, creating network bottlenecks. This paper addresses the need for an efficient, distributed strategy to balance energy consumption across all nodes in asynchronous WSNs, considering both individual node energy drain and the impact on neighboring nodes.

Methods

The authors first develop a mathematical model to characterize the energy efficiency of sensor nodes in asynchronous duty-cycling WSNs, specifically focusing on Low Power Listening (LPL) based protocols. They then formulate the Sleep Interval Control (SIC) problem as a convex optimization problem. To enable a practical, distributed solution, they decompose the global SIC problem into local sub-problems, leading to the Distributed Sleep Interval Control (D-SIC) problem.

The core of their contribution is the GDSIC (Generalized Distributed Sleep Interval Control) algorithm. This algorithm allows each sensor node to independently adjust its sleep interval by exchanging local information with its neighbors (receivers and senders). Nodes iteratively solve their local D-SIC problem to minimize their local maximum energy consumption rate. The algorithm is designed to be self-adjustable to traffic load variations and can be unified across various asynchronous MAC protocols. They also extend their approach to handle multi-receiver scenarios and receiver-initiated protocols.

Datasets

The paper uses a combination of:

  • Trace-driven simulations: They utilize a real trace harvested from the GreenOrbs WSN deployment, which provides data on network topology dynamics over a long period. This trace is filtered to mimic realistic link conditions.
  • Prototype testbed: A 50-node TelosB mote testbed is used for practical validation of the GDSIC algorithm.

Key Insights and Ideas

  • The "water-leveling" intuition is insufficient: Simply increasing the sleep interval of high-consumption nodes to match their neighbors doesn't guarantee global optimality. This can inadvertently increase the energy drain of their senders.
  • Asynchronous duty-cycling requires distributed control: Centralized solutions are impractical for large-scale WSNs due to communication overhead and scalability issues.
  • Convex optimization is key: Formulating the problem as a convex optimization problem allows for efficient decomposition and the development of distributed algorithms.
  • GDSIC balances local and global fairness: By focusing on local min-max energy fairness, the distributed algorithm iteratively converges towards a globally optimal solution.
  • Traffic awareness is crucial: The algorithm's ability to adapt to traffic load variance is essential for maintaining fairness in dynamic WSN environments.

Key Results

  • Improved Min-Max Energy Fairness: GDSIC significantly improves the min-max energy fairness compared to traditional identical sleep interval (EQUAL) and greedy strategies. In simulations, GDSIC achieved up to 64.1% improvement over EQUAL and 32% over GREEDY.
  • Balanced Energy Consumption: GDSIC leads to a much more balanced energy consumption rate across sensor nodes, reducing the likelihood of premature node failures. The difference between maximum and minimum energy consumption rates was reduced to 14.9% in GDSIC, compared to 48.9% in EQUAL.
  • Higher Network Yield: GDSIC achieved a higher network yield (percentage of reachable nodes) on average, by 50.8% and 35.9% compared to EQUAL and GREEDY, respectively, indicating better overall network connectivity and lifetime.
  • Practical Feasibility: Experiments on a 50-node testbed demonstrated the feasibility and applicability of GDSIC, showing convergence to stable energy consumption rates within minutes.

Limitations and Future Work

  • Initial Sleep Interval Setting: The paper acknowledges that determining an appropriate initial sleep interval can affect the efficiency of the SIC process and suggests this as a future research direction.
  • Leaf Node Handling: While addressed, the paper notes that leaf nodes (with no senders) might require specific considerations for optimal sleep interval adjustment.
  • Multi-receiver scenario extension: The paper outlines how GDSIC can be extended to multi-receiver scenarios by transforming it into an equivalent single-receiver case, but the detailed algorithm is omitted.
  • Receiver-initiated protocols: The paper mentions that the analysis and solution can be extended to receiver-initiated protocols like LPP, but the detailed rephrasing of energy consumption rates is not fully elaborated.

Summary updated

BibTeX
@article{li2014towards,
  author = {Zhenjiang Li and Mo Li and Yunhao Liu},
  title = {{Towards Energy-Fairness in Asynchronous Duty-Cycling Sensor Networks}},
  journal = {ACM Trans. Sens. Networks},
  year = {2014},
  doi = {10.1145/2490256},
}
Figure