Energy-driven occupant behaviour sensing

Samuel Chang Bing Wong, Stephanie Gauthier, Geoff V. MerrettePrints 2020

Abstract

Traditional occupant behaviour sensing (OBS) relies on invasive or high-maintenance battery-powered devices. This research explores Energy-Driven Systems (EDS) that harvest environmental energy to provide maintenance-free, privacy-respecting monitoring. By leveraging intermittent computing and transient networking, these systems can perform complex tasks like activity recognition with high accuracy despite frequent power interruptions.

Introduction

Optimizing building energy performance and occupant comfort requires detailed data on human movement and activity. However, current sensing technologies face two primary obstacles: privacy concerns related to cameras or microphones, and the logistical burden of maintaining large-scale battery-powered networks. The environmental impact and cost of replacing thousands of batteries make traditional wireless sensor networks unsustainable for long-term building management. Samuel C. B. Wong from the University of Southampton argues for a shift toward Energy-Driven Systems (EDS) to address these challenges.

The Energy-Driven Approach

Energy-Driven Systems (EDS) offer a solution by harvesting power from ambient sources such as light, airflow, or mechanical vibrations (e.g., footsteps). This paradigm shift removes the need for batteries but introduces the challenge of intermittency. Devices must operate in a state of constant power cycles, requiring a fundamental redesign of hardware and software architectures. The core philosophy is to design the application around the harvester's limitations rather than treating energy as an afterthought.

Technical Framework

The research proposes a cross-layer architecture designed around the limitations of energy harvesters, utilizing low-power microcontrollers like the MSP430 and specialized IoT protocol stacks like 6LoWPAN.

  • Intermittent Computing: To ensure computational progress, the system uses "checkpoints" to save the state to non-volatile memory before power is lost. This allows the processor to resume its task immediately upon "resurrection," ensuring forward progress in complex tasks.
  • Transient Networking: Communication in intermittent networks is difficult because internal clocks stop during power-down periods. The proposed framework utilizes wake-up radios and energy-aware routing protocols to manage connectivity despite frequent node dropouts and clock drift.

Results and Case Studies

A case study using the battery-less Wireless Identification and Sensing Platform (WISP) demonstrated that sparse data is sufficient for high-fidelity sensing. Using only accelerometer data from RF-powered tags, the system achieved near-perfect accuracy in activity recognition:

  • Lying: 100% precision and recall.
  • Sitting: 95-99% precision and recall.
  • Walking (Ambulating): 95-100% precision and recall.

Additionally, a proof-of-concept battery-less cycle computer was developed. It determines angular velocity based on the shape of power pulses generated by a magnet and coil, demonstrating that energy-driven sensing is viable in mechanical contexts where the energy source itself provides the data signal.

Conclusion

Energy-driven sensing represents a sustainable path forward for smart infrastructure. While intermittent computing has reached a level of maturity, significant work remains in scaling these systems to massive transient networks. Future research will focus on robust clock synchronization and the implementation of security protocols within the strict power envelopes of harvested energy, moving toward reliable, skyscraper-scale sensor networks.

Summary updated

BibTeX
@inproceedings{wong2020energy,
  author = {Samuel Chang Bing Wong and Stephanie Gauthier and Geoff V. Merrett},
  title = {{Energy-driven occupant behaviour sensing}},
  booktitle = {ePrints},
  year = {2020},
}
Figure