Sophisticated Batteryless Sensing

Wireless embedded sensing systems have revolutionized scientific, industrial, and consumer applications. Sensors have become a fixture in our daily lives, as well as the scientific and industrial communities by allowing continuous monitoring of people, wildlife, plants, buildings, roads and highways, pipelines, and countless other objects. Recently a new vision for sensing has emerged—known as the Internet-of-Things (IoT)—where trillions of devices invisibly sense, coordinate, and communicate to support our life and well being. However, the sheer scale of the IoT has presented serious problems for current sensing technologies— mainly, the unsustainable maintenance, ecological, and economic costs of recycling or disposing of trillions of batteries. This energy storage bottleneck has prevented massive deployments of tiny sensing devices at the edge of the IoT. This dissertation explores an alternative—leave the batteries behind, and harvest the energy required for sensing tasks from the environment the device is embedded in. These sensors can be made cheaper, smaller, and will last decades longer than their battery powered counterparts, making them a perfect fit for the requirements of the IoT. These sensors can be deployed where battery powered sensors cannot—embedded in concrete, shot into space, or even implanted in animals and people. However, these batteryless sensors may lose power at any point, with no warning, for unpredictable lengths of time. Programming, profiling, debugging, and building applications with these devices pose significant challenges. First, batteryless devices operate in unpredictable environments, where voltages vary and power failures can occur at any time—often devices are in failure for hours. Second, a device’s behavior effects the amount of energy they can harvest—meaning small changes in tasks can drastically change harvester efficiency. Third, the programming interfaces of batteryless devices are ill-defined and nonintuitive; most developers have trouble anticipating the problems inherent with an intermittent power supply. Finally, the lack of community, and a standard usable hardware platform have reduced the resources and prototyping ability of the developer. In this dissertation we present solutions to these challenges in the form of a tool for repeatable and realistic experimentation called Ekho, a reconfigurable hardware platform named Flicker, and a language and runtime for timely execution of intermittent programs called Mayfly.