[ad_1]
The United States produces about 68 million tonnes of recycling each year (a figure that, frankly, should be much higher).
It's a lot of paper, plastic, metal and glbad (and sometimes real waste) that human workers can filter and sort.
It's monotonous, dirty and often dangerous work, but someone needs to do it. So why not a robot?
Researchers at MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) have developed a robotic system that can detect whether an object is made of paper, metal or plastic.
The soft teflon hand of the "RoCycle" system uses touch sensors at your fingertips to detect the size and rigidity of an object. Basically, he squeezes cups, boxes and cans to determine their composition.
No visual cue needed.
"The sensed skin of our robot provides a haptic feedback that allows it to differentiate a wide range of objects, from rigid objects to soft ones," said Daniela Rus, lead author of the study and professor of MIT in a statement.
In collaboration with Yale University, RoCycle demonstrates the limitations of sight sorting. it can distinguish between two identical Starbucks cups, made of paper and plastic, that can cause problems for vision systems (and the human eye).
"Computer vision alone can not solve the problem of giving machines a human perception," Rus said. "So being able to use a touch input is of vital importance."
For decades, the United States has sent most of its recycling to China, converting it into shoes, bags and other new products. But last year, the country raised its standards by restricting imports of blended paper and most plastics, which means that a majority of our single-stream recycling business now ends up in landfills. .
"If a system such as RoCycle could be deployed on a large scale, we could potentially benefit from the convenience of recycling in a single stream with lower contamination rates of multi-stream recycling," says PhD student Lillian Chin, lead author. on the MIT paper.
RoCycle is a robot capable of sorting recycling (via Jason Dorfman / MIT CSAIL)
Designing a machine to do our dirty work, however, is not easy. While humans can hold an object and immediately recognize many of its qualities, even with their eyes closed, the robot's hands are not yet as advanced.
This is where Rus & Co. come in.
Using a motorized hand made of a relatively new material called "auxetic", the team created a component that, when cut, twists to allow dynamic movement.
An badociated set of sensors helps to estimate the size and rigidity of an element, able to differentiate between "hard" objects and "soft" objects with an accuracy of 78%.
Compatible with all robotic arms, RoCycle has detected an accuracy of 85% for material detection at shutdown and 63% on a real simulated transport belt.
According to MIT, the most common mistake was to identify paper-covered metal boxes, which the team said would be improved by adding more sensors along the contact surface.
More on Geek.com:
[ad_2]
Source link
