Abstract
The Internet-of-Things holds the promise of realizing ubiquitous computing in its full potential. Sensors that work as the fundamental building blocks of the IoT have become an integral part of our everyday lives. They sense, compute, and communicate to monitor humans, pets, wildlife, marine life, plants, crops, buildings, factories, city infrastructures, and many others. As the network of computing devices continues to grow rampantly, in one or two decades, there will be a hundred sensors per person on earth. At this scale, sensors must be long-lived to curtail the intractable cost of maintenance and the negative environmental impact caused by short-lived batteries and outdated electronics.
This dissertation argues and establishes that perpetually-powered energy-harvesting devices, instead of battery-powered ones, are the key to enforcing a sustainable Internet-of-Things. Self-powered devices are perpetual, zero-maintenance, eco-friendly, and pervasively deployable. Together with sustainable power, we emphasize utilizing devices that are already installed in place to enable long-lasting design points through retrofitting and repurposing. However, the energy intermittency inherent in batteryless power supplies imposes two major challenges that limit the adoptability of energy-harvesting sensors: complexity in application design and highly unreliable service quality. To overcome these challenges, we introduce an energy supervisor architecture named ALTAIR, which abstracts the details of energy management from application software to simplify batteryless designs. Moreover, we propose PreFarad, a system architecture that isolates and prioritizes the sensor’s energy requirement from the rest of the system components to improve the event detection accuracy of intermittent sensors. Additionally, we extend the functionality of an energy-harvesting power supply to enable two sustainable design points by incorporating new sensing capabilities on existing devices. RETROIOT upgrades existing IoT devices with additional sensing features as well as an energy-harvesting power supply. Lastly, we demonstrate SolarWalk design point that transforms a photovoltaic energy-harvester to an accurate sensor. These systems significantly enhance the capabilities of today’s energy-harvesting batteryless IoT sensors.
