Dewdrop: an energy-aware runtime for computational RFID

Computational RFID (CRFID) tags embed sensing and computation into the physical world. The operation of the tags is limited by the RF energy that can be harvested from a nearby power source. We present a CRFID run-time, Dewdrop, that makes effective use of the harvested energy. Dewdrop treats iterative tasks as a scheduling problem to balance task demands with available energy, both of which vary over time. It adapts the start time of the next task iteration to consistently run well over a range of distances between tags and a power source, for different numbers of tags in the vicinity, and for light and heavy tasks. We have implemented Dewdrop on top of the WISP CRFID tag. Our experiments show that, compared to normal WISP operation, Dewdrop doubles the operating range for heavy tasks and significantly increases the task rate for tags receiving the least energy, all without decreasing the rate in other situations. Using offline testing, we find that Dewdrop runs tasks at better than 90% of the best rate possible.