An Energy Aware Adaptive Sampling Algorithm for Energy Harvesting WSN with Energy Hungry Sensors

Bruno Srbinovski, Michele Magno, Fiona Edwards Murphy, Vikram Pakrashi, Emanuel M. PopoviciSensors 2016 · DOI

TL;DR

This paper introduces an "Energy Aware Adaptive Sampling Algorithm" (EASA) for wireless sensor networks (WSNs) that are equipped with energy harvesting capabilities and power-hungry sensors. EASA dynamically adjusts the sensor sampling rate based on the available energy, aiming to achieve self-sustainability and extend network lifetime, outperforming traditional adaptive sampling methods.

What Problem Does This Paper Address?

Wireless Sensor Networks (WSNs) are crucial for long-term monitoring, but their limited battery life is a major bottleneck. While energy harvesting offers a solution, the intermittent and unpredictable nature of harvested energy, coupled with the high power demands of modern sensors (especially those requiring warm-up periods or frequent sampling), makes achieving true self-sustainability challenging. Existing adaptive sampling algorithms reduce sampling frequency based on signal dynamics but don't explicitly account for the node's energy reserves, potentially leading to premature battery depletion. This paper addresses the need for an energy management strategy that integrates both energy harvesting and adaptive sampling to ensure continuous operation of WSNs with power-hungry sensors.

Methods

The core of this work is the proposed Energy Aware Adaptive Sampling (EASA) algorithm. EASA builds upon a traditional Adaptive Sampling Algorithm (ASA) by incorporating an energy awareness module.

Here's how it works:

  • ASA Foundation: The base ASA monitors the signal's maximum frequency (fMax) and adjusts the sampling frequency (fsampling) to be at least twice fMax to avoid aliasing. This is a standard technique to reduce unnecessary sampling when the signal is stable.
  • Energy Awareness Module: EASA introduces a critical battery level (Xlevel). When the battery level (Ebatt) is above Xlevel, EASA functions like a standard ASA, using the sampling rate dictated by the signal dynamics. However, when Ebatt drops below Xlevel, EASA becomes more conservative. It reduces the sampling frequency (fEASA) based on a formula that considers the current battery level and a parameter 'm' (which controls the rate of reduction).
  • Energy Model: A detailed energy model for an energy harvesting WSN node with a power-hungry sensor is developed. This model accounts for various operational states (sleep, receive, transmit, sensor sampling, processing) and their respective power consumptions and durations. It also incorporates harvested energy from different sources.
  • Simulation and Evaluation: The EASA algorithm and the energy model were implemented in MATLAB. The performance was evaluated using in-field measured data from two distinct deployments: one with wind energy harvesting and ultrasonic wind sensors, and another with solar energy harvesting and gas sensors in a beehive. Simulations compared EASA against a traditional ASA and a fixed sampling rate approach.

Datasets

The paper utilizes real-world data collected from two distinct deployments:

  • Deployment (I): Wind speed and direction data collected over 35 days from a custom-built platform using ultrasonic wind sensors and a micro wind generator for energy harvesting.
  • Deployment (II): Data collected over 14 days from an off-the-shelf WSN platform (Libelium "Waspmote") monitoring gas levels (CO2, O2, NO2, etc.) in a beehive, using solar energy harvesting.

These datasets were used to characterize sensor power consumption, energy harvesting profiles, and signal dynamics for simulation and evaluation of the EASA algorithm.

Key Insights and Ideas

  • Energy Awareness is Crucial for Self-Sustainability: Simply adapting sampling based on signal dynamics isn't enough for energy harvesting WSNs. Explicitly linking sampling rate to the node's energy reserves is key to achieving long-term, self-sustainable operation.
  • Power-Hungry Sensors are a Major Challenge: Sensors that require significant warm-up times or prolonged sampling periods are particularly problematic for battery-powered WSNs. EASA's ability to dynamically reduce sampling frequency is vital for managing their energy demands.
  • Trade-off Between Accuracy and Longevity: EASA's energy conservation strategy can lead to a reduced sampling rate that might fall below the Nyquist criterion, potentially introducing aliasing and measurement errors. The paper acknowledges this as a trade-off for extended network activity, suggesting that some errors might be acceptable or correctable through post-processing.
  • Generic Applicability: The proposed EASA and its energy model are designed to be generic, allowing implementation on various WSN platforms and with different energy harvesting sources and power-hungry sensors.

Key Results

  • Significant Lifetime Extension: Simulations using in-field data demonstrated that EASA can drastically increase the lifetime of a WSN node compared to traditional ASA or fixed sampling rate methods, enabling self-sustainability.
  • Outperformance of ASA: In scenarios where energy levels drop below a critical threshold, EASA significantly reduces sampling rates, leading to substantial energy savings compared to a standard ASA that continues to sample based solely on signal dynamics.
  • Validation Across Diverse Scenarios: The algorithm was successfully evaluated in two distinct application domains (wind monitoring and beehive monitoring) with different sensor types and energy harvesting sources (wind and solar), showcasing its versatility.
  • Error vs. Activity Trade-off: In cases where the energy level is critical and the parameter 'm' is set to a more conservative value (e.g., 1/3), the sampling rate can drop below the signal's maximum frequency, leading to potential aliasing. However, this is presented as a necessary compromise to keep the node active.

Limitations and Future Work

  • Aliasing and Measurement Error: The paper acknowledges that when the battery level is critical, EASA might reduce the sampling rate to a point where it violates the Nyquist criterion, potentially introducing aliasing and measurement errors. This is presented as a trade-off for staying active.
  • Parameter Tuning: The optimal values for parameters like Xlevel (critical battery level) and 'm' (energy conservation rate) are not explicitly determined and would likely depend on the specific application and sensor characteristics.
  • Future Work: The authors suggest that further discussion on filtering aliased signals and accepting measurement errors as a trade-off for network activity is warranted. Additionally, exploring more sophisticated energy management strategies and optimizing parameter selection for different scenarios would be valuable.

Summary updated

BibTeX
@article{srbinovski2016energy,
  author = {Bruno Srbinovski and Michele Magno and Fiona Edwards Murphy and Vikram Pakrashi and Emanuel M. Popovici},
  title = {{An Energy Aware Adaptive Sampling Algorithm for Energy Harvesting WSN with Energy Hungry Sensors}},
  journal = {Sensors},
  year = {2016},
  doi = {10.3390/s16040448},
}
Figure