The objective of this research is to develop the circuitry and software to enable postage stamp sized(or smaller) energy harvesting, batteryless sensing devices that adapt their function to long and shortterm changes in their environment—providing a minimum quality of application regardless of energyavailable. This adaptation is both necessary, and difficult. Necessary because these devices promisenew applications in essential areas—health and wellness wearables and implants, infrastructuremonitoring, and small animal tracking. Difficult because of the intermittent execution caused byunpredictable energy harvesting conditions and environments these devices must operate in. Byforgoing expensive and short-lived traditional energy storage (batteries) these devices cantheoretically be deployed in vast numbers at low cost, maintenance-free, for decades. However, theselong, massive deployments have not yet been attempted. By providing a way to approximate andscale tasks based on predicted energy available, and adapt these tasks to changes in availableenvironmental energy, the proposed research will allow these tiny devices to be deployed confidently,enable long maintenance free lifetimes, and finally allow for large scale deployments.The techniques developed from this proposal will address the challenge of adaptation in volatileenvironments for intermittently powered energy harvesting devices in two interconnected parts: (1)in the form of a co-designed hardware / software environmental profiler, and (2) an approximatingtask based runtime and in-situ environment recognition circuit. The profiler tool will gather realisticenergy harvesting conditions as well as metadata in the actual deployment location, then generate adeployment profile from the data that selects commonly encountered environmental conditions andtransitions (day to night, seasonal, even body movement) that can be keyed for adaptation. Thisinformation informs the task based runtime, which will execute tasks approximately based on theenergy environment conditions identified by the recognition circuit, and programmer definedconstructs. Together, these tools make energy harvesting sensors more practical, usable, andefficient, enabling applications across health, infrastructure, military, and commercial domains.
|Effective start/end date||2/15/19 → 1/31/22|
- National Science Foundation (CNS-1850496-001)