In municipal recycling facilities, workers pick through discarded items going past on a conveyor belt, sorting them according to their material. While this setup is a source of employment, it's also costly, limiting the economic feasibility of such operations. A new material-detecting robot hand, however, could help bring costs down.
Known as RoCycle, the device is being developed in a partnership between MIT and Yale University. Compatible with any third-party robotic arm, it's reportedly 85 percent accurate at differentiating between paper, plastic and metal items when they're sitting still, and 63 percent accurate when they have to be grabbed off a moving conveyor belt. Both of those figures should rise as the technology is developed further.
RoCycle utilizes two soft-bodied fingers to pick up objects, gently squeezing them as it does so. First of all, this allows it gauge the size of each item. Additionally, utilizing pressure sensors in the fingers, it measures the amount of force that's required to grasp the object. This indicates the stiffness of that object's material.
Custom software then compares those size and stiffness figures to those for a variety of known paper and plastic items, that have previously been entered into a database. When a match is found, the system knows what the hand is holding, and what that thing is made of. The pressure sensors are also conductive, which lets the system identify metal objects.
Once an item's material has been ascertained, the RoCycle-equipped arm proceeds to drop the object in an appropriate bin.
And while the fingers are soft and squishy, they're not as fragile as other "soft robotic" devices. That's because instead of having a balloon-like body, each finger consists of two parallel cylinders, covered in a silicone sleeve. Those cylinders are in turn made of an auxetic polymer, which actually gets wider as it's stretched. When a motor turns the cylinders, the polymer twists and stretches, thus getting wider and making the finger fatter.
Called "handed shearing auxetics" (HSA), this technology not only makes for soft fingers that are still relatively hardy, but it also allows regular motors to be used. By contrast, some fluid-based soft robotic systems require more expensive air pumps or compressors.
The researchers are now working on combining the RoCycle tech with video output from robots' cameras, allowing for greater accuracy by using both tactile and visual data. "Computer vision alone will not be able to solve the problem of giving machines human-like perception, so being able to use tactile input is of vital importance," says MIT's Prof. Daniela Rus, lead author of a paper on the research.
That paper is being presented later this month at the IEEE International Conference on Soft Robotics in Seoul, South Korea. In the meantime, RoCycle can be seen in action, in the video below.
Source: MIT Computer Science and Artificial Intelligence Lab (CSAIL)
