Harvester-aware transient computing: Utilizing the mechanical inertia of kinetic energy harvesters for a proactive frequency-based power loss detection

摘要

Power-neutral system design avoids energy buffers by directly powering the load by the energy harvester. In case of a power loss, checkpointing methods ensure forward progress by preserving the volatile system state using non-volatile memories. The timely detection of upcoming power losses is essential for a reliable checkpointing process. Moreover, various applications require early detections to, e.g., ensure the finalization of atomic operations. However, common voltage threshold-based methods only allow short-term detections.In this paper we propose a new methodology that allows early detections by exploiting physical characteristics of the harvester. To this end, small-scale kinetic energy harvesters are considered that employ rotatably mounted mechanical masses to drive electromagnetic generators. Due to the inertia of these masses, the power output does not stop abruptly, but gradually decays after the excitation of the harvester is over.We investigate the relationship between the initial excitation intensity as it is reflected in the output frequency, the load current and the remaining period of power availability. Our results indicate that this relationship allows to predict the power duration based on the output frequency of the harvester. We show that power losses can be detected up to one order of magnitude earlier with our frequency-based method than with state-of-the-art voltage-based methods.