Energy Efficient Data Collection and Directional Wireless Power Transfer in Rechargeable Sensor Networks

Zhigang Duan, Lei Tao, Xin Ming ZhangIEEE Access 2019 · DOI

Abstract

Wireless Power Transfer (WPT) offers a sustainable solution for powering rechargeable sensor networks (RSNs), yet traditional omnidirectional methods suffer from significant energy waste. This paper proposes a Directional Wireless Power Transfer (DWPT) framework that jointly optimizes energy beamforming and data collection. By employing a sector-based antenna model and a machine learning-based optimization shortcut, the proposed approach achieves superior network utility and energy efficiency compared to conventional methods.

Introduction

The proliferation of the Internet of Things (IoT) has led to the deployment of vast sensor networks. Maintaining these networks often involves the costly and labor-intensive task of battery replacement. Wireless Power Transfer (WPT) has emerged as a promising alternative, enabling sensors to harvest energy from dedicated Power Beacons (PBs). However, standard omnidirectional WPT broadcasts energy indiscriminately, leading to low efficiency in sparse networks. This paper explores Directional Wireless Power Transfer (DWPT) to focus energy where it is needed most.

System Model and Problem Formulation

The proposed system operates in discrete time frames. Each frame is divided into an Energy Transfer (ET) phase and an Information Transfer (IT) phase. During the ET phase, PBs transmit energy to sensors; during the IT phase, sensors use that energy to transmit collected data to a sink.

Figure 2: The time frame structure consisting of Energy Transfer (ET) and Information Transfer (IT) phases.

The core challenge lies in the joint optimization of PB transmit power, directional antenna gains, and sensor data rates to maximize network utility while ensuring energy neutrality (sensors don't run out of power) and buffer stability. This results in a complex, non-convex optimization problem.

Proposed Methodology

To address the optimization challenge, the authors introduce several key components:

Sector-Based Antenna Model

PBs utilize directional antennas modeled by dividing the surrounding space into $M$ sectors. One sector is designated as the main lobe with high gain, while others receive lower side-lobe gains. This simplification allows for more tractable optimization of energy beamforming.

Optimization and Machine Learning

The authors employ Block Coordinate Descent (BCD) to iteratively solve the joint optimization problem by breaking it into sub-problems. Recognizing the computational constraints of sensor nodes, they also propose a k-Nearest Neighborhood (k-NN) framework. By training the k-NN model offline using BCD-generated data, the system can rapidly predict optimal antenna directions based on real-time sensor positions.

Figure 5: Network topology illustrating Power Beacons (PBS), Sensor Nodes (SN), and the Sink within a sectorized environment.

Results and Discussion

Simulations in random network topologies demonstrate that the DWPT scheme consistently outperforms omnidirectional baselines. A key finding is the delay-efficiency trade-off: increasing the allowable delay (the time sensors can wait to harvest energy) significantly boosts energy efficiency in the directional model. Furthermore, the results show that adding more Power Beacons provides a more substantial efficiency gain than increasing the number of sensors, highlighting the importance of spatial diversity in energy harvesting.

Conclusion

By transitioning from omnidirectional to directional wireless power transfer and integrating machine learning for real-time optimization, RSNs can achieve significantly higher energy efficiency and data utility. Future work will focus on extending this model to dynamic topologies where sensors and beacons are mobile.

Summary updated

BibTeX
@article{duan2019energy,
  author = {Zhigang Duan and Lei Tao and Xin Ming Zhang},
  title = {{Energy Efficient Data Collection and Directional Wireless Power Transfer in Rechargeable Sensor Networks}},
  journal = {IEEE Access},
  year = {2019},
  doi = {10.1109/access.2019.2959060},
}
Figure