Researchers at Carnegie Mellon University's Future Interfaces Group have developed self-powered radio tags, dubbed Sozu, which allow for building-scale activity sensing without the need for line-of-sight and at a very low cost — complete, they claim, with a 99 percent detection rate and "almost no false positives."
The brainchild of researchers Yang Zhang, Yasha Iravantchi, Haojian Jin, Swarun Kumar, Chris Harrison, Sozu is designed to offer robust wide-area sensing of activity in a human-populated environment at a low cost. To do that, the team started by dropping the idea of having a battery or mains power connection for the nodes themselves. "Sozu tags convert energy from activities that they sense into RF broadcasts," the team explains in their paper on the subject, "acting like miniature self-powered radio stations.
"A Sozu deployment consists of one antenna, which can be placed in an inconspicuous location, such as a basement. Users then attach Sozu 'tags' to items and infrastructure of interest. To keep cost low, these tags are constructed from ultra-low-cost analogue components, and thus cost only a few dollars each (i.e., no digital components, nor digital communication like Wi-Fi or Bluetooth).
While similar in concept to radio-frequency ID (RFID) tags, which harvest energy from a nearby transmitter to power a reply, Sozu harvests energy from a different source: the monitored activity itself. "Another key innovation is the battery-less design of Sozu tags, which instead harvest energy directly from the activities they sense," the team notes. "The tags convert and then broadcast this energy as radio frequency (RF) waves, acting like small radio stations. Each tag is given a unique frequency, allowing for recognition of many simultaneous events. Importantly, we selected a frequency range that readily penetrates common building construction types, offering whole-building sensing with just a single antenna."
The types of energy able to be harvested by the tags are impressive. Sozu implementations were created that convert motion energy, vibration energy, solar energy, thermal energy, electromagnetic radiation, and water flow into enough power to transmit an activity.
The team worked with eight Human-Computer Interaction students to test the system, which they claim can easily and cheaply scale to whole-building monitoring. After a brief introduction, the students implemented Sozu-based monitoring systems for parking occupancy, medication reminders triggered if the pill bottle lid hasn't been opened, a smart recycling bin, an automatic door for wheelchairs, consumables monitoring, work-hour monitoring, meeting room occupancy, and foot traffic monitoring. All were successful, while detailed testing revealed that 99 percent of activities were successfully recorded with a very low rate of false positives.
The team's paper, which was presented at the 32nd Annual ACM Symposium on User Interface Software and Technology (UIST '19), is available here.