UW News

September 22, 2014

New RFID technology helps robots find household objects

Mobile robots could be much more useful in homes if they could accurately locate people, places and objects. Today’s robots usually see the world with cameras and lasers, which have difficulty reliably recognizing objects and can miss those hidden in clutter.

A PR2 robot successfully navigates to a medication bottle.

A PR2 robot successfully navigates to a medication bottle.

A complementary way robots can “sense” what is around them is through the use of small radio-frequency identification, or RFID, tags tuned to ultra-high frequencies. Inexpensive, self-adhesive tags can be stuck on objects, allowing an RFID-equipped robot to search a room for the correct tag’s signal. Once the tag is detected, the robot knows the object isn’t far away.

“But RFID doesn’t tell the robot where it is,” said Charlie Kemp, a professor at Georgia Institute of Technology. “To actually find the object and get close to it, the robot has to be more clever.”

The University of Washington’s Matt Reynolds, an associate professor of electrical engineering and of computer science and engineering, teamed with Kemp and former Georgia Tech student Travis Deyle to develop a new search algorithm that improves a robot’s ability to find and navigate to tagged objects. The team has implemented its system on a type of robot called a PR2, allowing it to travel through a home and correctly locate different types of tagged household objects, including a medication bottle, TV remote, phone and hair brush.

The research was presented this month at the IEEE/RSJ International Conference on Intelligent Robots and Systems in Chicago.

The researchers equipped a PR2 robot with articulated, directionally sensitive antennas and a new algorithm that allows the robot to successfully find and navigate to an object. These antennas tend to receive stronger signals from a tag when they are closer to it and pointed more directly at it.

PR2 robotBy moving the antennas around on its shoulders and driving around the room, the robot can figure out the direction it should move to get a stronger signal from a tag and thus become closer to a tagged object.

In essence, the robot plays the classic childhood game of “Hotter/Colder,” with the tag telling the robot when it’s getting closer to the target object.

“While we have demonstrated this technology with a few common household objects, the RFID tags can uniquely identify billions of different objects with essentially zero false positives. This is important because many objects look alike, yet must be uniquely identified,” Reynolds said.

In contrast to other approaches, the robot doesn’t explicitly estimate the 3-D location of the target object, which significantly reduces the complexity of the algorithm.

“Instead, the robot can use its mobility and our special behaviors to get close to a tag and orient toward it,” said Deyle, who conducted the study in Kemp’s lab while earning his doctoral degree.

Deyle, who currently works at Google Inc., says the research has implications for future home robots and is particularly compelling for applications like helping people take medicine.

“This could allow a robot to search for, grasp and deliver the right medication to the right person at the right time,” Deyle said. “RFID provides precise identification, so the risk of delivering the wrong medication is dramatically reduced. Creating a system that allows robots to accurately locate the correct tag is an important first step.”

This research was funded by the National Science Foundation and the Willow Garage PR2 Beta Program.


For more information, contact Reynolds at matt.reynolds@ee.washington.edu.

This story was adapted from a Georgia Institute of Technology news release.