MIT researchers have developed a system that can efficiently retrieve any object buried in a pile.
The breakthrough builds on previous research that combined visual information and radio frequency (RF) to enable a robotic arm to find objects marked with RFID tags, which reflect signals sent by an antenna.
A robot requires complex reasoning to complete such a challenge.
Named FuseBot, the algorithm controlling the system reasons about the probable location of objects within the pile.
This is achieved by using FuseBot, a robotic arm with a video camera attached and an RF antenna to retrieve an untagged target from a mixed pile.
According to MIT, the system simultaneously scans the pile to create a 3D model of the environment to find the space the object is likely to occupy and sends signals from the antenna to find RFID tags.
MIT claims these radio waves can pass through most solid surfaces, giving the robot good depth of vision into the pile. The target is not tagged, meaning FuseBot knows the item cannot be located in the same place as an RFID tag.
Next, FuseBot is designed to search for the most efficient way to remove obstructions and extract the target object.
The robot also uses pressing to determine the texture of objects.
After choosing an object to discard, the robot re-scans the pile and uses strategic reasoning to decide which object to remove next.
MIT claims that it is this reasoning that reduces the time taken to find objects in half when compared to a ‘state-of-the-art robotics system’ that uses only vision.
Fadel Adib, senior author and associate professor in the Department of Electrical Engineering and Computer Science and director of the Signal Kinetics group in the Media Lab at MIT, said: “This speed could be especially useful in an e-commerce warehouse.
“A robot tasked with processing returns could find items in an unsorted pile more efficiently with the FuseBot system.
“What this paper shows, for the first time, is that the mere presence of an RFID-tagged item in the environment makes it much easier for you to achieve other tasks in a more efficient manner.
“We were able to do this because we added multimodal reasoning to the system – FuseBot can reason about both vision and RF to understand a pile of items.”
MIT ran more than 180 trials involving a range of household items, including office supplies, stuffed animals and clothing. Pile sizes and the number and size of RID-tagged items varied each time.
FuseBot is said to have successfully located the target 95% of the time and in 40% fewer moves than the other system.
