Backscatter Protocols and Energy-Efficient Computing for RF-Powered Devices
MetadataShow full item record
RF-powered computers operate using energy they harvest from radio frequency (RF) signals. Compared to battery powered devices, they have the advantages of being small and long-lived as they do not need to carry an onboard energy-store, and they can be embedded inside objects, structures and even the human body. In this dissertation, I explore how we can build networks of RF-powered computers that support rich functionality across a range of RF environments. Towards this goal, I identify the key challenges to running programs on RF-powered computers, and argue that devices must adapt their behavior to match their task energy consumption to the available power. To demonstrate this approach, I formulate the task scheduling problem for one class of RF-powered devices (computational-RFIDs), and implement Dewdrop, an energy-aware CRFID runtime. By waking tags at the right times, Dewdrop can complete tasks that previously could not complete, and at close to their maximum rate given the power that the RF environment provides. The second challenge I tackle is how to build networks using backscatter communica- tion. Backscatter signaling is an ultra-low power form of communication, but the simplicity needed to achieve such low power operation makes clients prone to interference. To under- stand how this challenge can be overcome, I use measurement and simulation results to ex- plore the mechanisms by which interference impacts network performance. I then use these insights to develop a network design that mitigates interference and enables backscatter networks to scale well. Experimental results show that this design improves both coverage and capacity in a building-scale network compared to existing designs. This dissertation supports my thesis that RF-powered computers can support rich tasks in a variety of RF environments, and networks of such devices can scale to building-sized deployments. As technology advances, RF-powered devices will only decrease in size and increase in computational power and operational range. By demonstrating that they can also support rich functionality and be used to build building-scale networks, this disserta- tion demonstrates their potential to provide deeply embedded and long-lived sensing and computation.