In the cold of January, researchers dragged a cart with a train of 50 RFID tags across the 10th Street Bridge in Pittsburgh.
Electrical and Computer Engineering (ECE) Assistant Professor Swarun Kumar and School of Computer Science Professor Jason Hong along with students Haojian Jin, Jingxian Wang, and Zhijian Yang conducted the frigid experiment to measure the bridge’s curvature. A computer readout reconstructed the bridge’s shape so accurately that it matched infrastructure surveys. The team is creating a shape-aware technology, called WiSh, and they plan to apply it to anything with a shape.
Wang, Ph.D. student in ECE, says that designs of smart cities served as an inspiration to create WiSh. “Low-cost, low-power, and minimum-effort are the main concerns when people design smart cities,” he said. Infrastructure monitoring has so far hit hard limits in China and the US. Many of these monitoring systems rely on expensive and energy-intensive sensors such as cameras or motion sensors that create more problems than they solve. WiSh could operate in cities cheaply, accurately, and battery-free, sensing a diverse range of buildings and shapes.
Beyond applications in infrastructure, WiSh in smart clothing could measure movement and posture, such as spine curvature and gestures. VR gaming could benefit from WiSh through smart clothing and tags embedded in carpeting or upholstery to more accurately capture a player’s movement.
“WiSh is inspired by our recent body-tracking work, RFWear, and by industrial products, like vision-based shape-aware technology that Disney and Microsoft have explored as motion capture for virtual reality,” Wang said. WiSh especially has key advantages over motion capture: cost, ease of use, and power supply.
“A lot of daily-wearable tech devices are costly, require constant charging, and can’t be washed,” said Jin, a Ph.D. student in the Human-Computer Interaction Institute. “With WiSh, its RFID tags cost only a cent apiece, are easy to implement, and require no batteries.”
WiSh works by analyzing the RFID signals to model the shape of a surface. The tags on the bridge, for instance, appear as a curve. In most devices using RFID tags, such as ID cards and readers, the tags are mobile, and the reader, stationary. The team reversed this to make the tags stationary and the reader mobile. This enables the team to place tags on any surface and position an RFID reader in any desired location.
But to capture the positions of each tag requires a great deal of computing power. Otherwise, a computer would struggle to model more than a few dozen tags. The team created algorithms that matched the positions of tags as a whole to real-world models. This approach drastically reduces computing time and nullifies information gaps from unaccounted-for tags.
The team, advised by Kumar, discovered some limitations in how they can use the RFID tags. Because the tags themselves are 3cm2, the algorithm can’t resolve any features on a surface smaller than that, such as wrinkles on cloth. Similarly, the algorithm can’t detect folds because of the RFID tags being placed on and coupling with each other, thus obfuscating a clear read.
While the RFID tags are currently being used on the City of Pittsburgh bridges, the team expects the novel devices to be broadly applicable in the future.
“From smart cities and homes, to every day materials like gaming systems and clothing,” remarked Wang. “These tags will be life-changing.”
Learn more about Swarun Kumar's research: