MIT Builds Drone-Based RFID Relay to Track Boxes in Warehouses

A silver drone flies through an aisle of sheleves lined with boxes in a warehouse.
Photo: MIT

Boxes are everywhere: The United States Postal Service shipped 5.2 billion packages in 2016, while Amazon shipped 1 billion packages just for the holidays. A single Girl Scout has sold an astounding 101,100 boxes of cookies.

Keeping track of all these boxes is a real pain. Retailers, shipping companies, and manufacturers are always looking for better ways to manage inventory. In recent years, they’ve begun using RFID tags to track boxes from warehouses to trucks to retail stores and then to customers’ homes.  

But RFID tags aren’t perfect—at every step, employees must scan each tag with an RFID reader. These scans must be done at close range, because RFID tags harness power from the signal the reader transmits to them.

Recently, researchers have begun to investigate other ways for these industries to trace items, including systems that can snap photos of labels or identify shipments by other visual cues. Now, an MIT group thinks they have an even better answer.

A team led by Fadel Adib, a principal investigator at MIT Media Lab, attached an RFID relay to a drone that he says could fly around and scan all the RFID tags on every box in a warehouse, and transmit that information back to a reader.

Using their prototype, the RFID tags and reader can be placed up to 50 meters apart. The system can also correctly record the location of a box to within a meter at the 50-meter range. At shorter ranges, it can pinpoint a box’s location to within 19 centimeters.

In telecommunications, relays are commonly used to boost signals from a cell tower to reach customers on the edge of a service area. But the MIT group is the first to apply this concept to RFID tags and execute it with a drone. Eventually, Adib wants to create a daisy chain of drones that can relay information to one another and forward it to multiple readers.

For now, the test system consists of a single drone that forwards a signal from an RFID reader to a tag. The tag uses some of the energy from that signal to power itself, and also encodes its identifier on the signal before sending it back. The drone then forwards that signal on to the reader, which decodes the identifier.

In addition to knowing which boxes are in a warehouse, the MIT team also wanted their system to record the location of each box. “We were wondering whether we could use drones to find missing items,” Adib says.

This presented another challenge. One way to do this would be to use an array of multiple antennas. When a signal is sent to the array, each antenna receives it from a slightly different angle. Software can then triangulate the source’s location from these angles.

But antenna arrays are too bulky for a drone to carry, or at least the type of drone that could safely be flown indoors. So, the group came up with another solution—using the drone’s flight path as a reference point to work out the tag’s location.

In their system, a tag transmits several signals in quick succession to the drone, and those signals arrive at different angles because the drone is flying around. This makes it possible for software to compare the angle of arrival from multiple signals, similar to how an antenna array records multiple angles of the arrival for one signal.

However, they soon ran up against another problem—the drone itself moves relative to the reader. Since all the processing to figure out the tag’s location is handled by software linked to the reader, the team needed a way to factor in the drone’s location when it took its angle-of-arrival measurements. The group decided to attach an RFID tag to the drone, so the software could find the drone’s position when it was working out each tag’s location.

For the project, the group used a US $499 Parrot BEBOP 2 drone and Alien Squiggle passive RFID tags, which cost 12 cents apiece. They described their system at the Association for Computing Machinery's Special Interest Group on Data Communications in August.

The team had initially tried to attach an RFID reader directly to a drone, but the reader was too heavy. “We tried that, and the drone crashed,” Adib says. “We had to buy another drone.”

Their relay weighs only 35 grams, compared to the lightest RFID readers, which weigh 500 grams. They powered the relay with the drone’s battery; it consumed about 3 percent of the battery’s power. This would reduce the drone’s 30-minute flight time by 2 or 3 minutes.

An RFID reader, and their drone relay, can process up to 100 tags per second. And unlike camera-based systems, it can zoom by and log a box even if the tag itself isn’t in plain sight.

For now, someone must fly the drone, but Adib would like to develop software to pilot these drones autonomously. His group is working with a retailer in Massachusetts to test their original system in a warehouse. “I imagine that in the future, you could have a fleet of drones that can, on their own, fly over and catalog all of these items,” Adib says.



IEEE Spectrum’s award-winning robotics blog, featuring news, articles, and videos on robots, humanoids, drones, automation, artificial intelligence, and more.
Contact us:

Erico Guizzo
New York City
Senior Writer
Evan Ackerman
Washington, D.C.

Newsletter Sign Up

Sign up for the Automaton newsletter and get biweekly updates about robotics, automation, and AI, all delivered directly to your inbox.