Recycling robotics uses AI to sort food and beverage cartons from other materials
The AMP Cortex system was created by AMP Robotics, developed in collaboration with the Carton Council of North America.
The Carton Council of North America was formed in 2009 by four major carton manufacturers: Elopak, SIG Combibloc, Evergreen Packaging and Tetra Pak, with the shared goal of making carton recycling more efficient for facilities and accessible for consumers.
Earlier this year, the Carton Council announced it had achieved 60% carton recycling access for US homes, a significant increase from 18% household access in 2009.
And now, through a pilot program using the AMP Cortex at Alpine Waste & Recycling in Denver, Colorado, AMP Robotics and the Carton Council say they have improved the efficiency and accuracy of carton recycling even further.
Consumers looking for renewable packaging
Renewable carton packaging has been increasingly popular for food and beverage products because of its low carbon footprint and lightweight structure.
However, with increased use came a need to divert the material away from landfills.
According to Nielsen, 63% of consumers said renewable packaging is a key driver in purchasing decisions, but more recyclable products meant more packaging ending up at recycling facilities in need of sorting.
Quicker than the status quo
The AMP Cortex has “spider-like” arms controlled by artificial intelligence (AI) to identify, grab, and separate food and beverage cartons from a line at a material recovery facility (MRF).
“We’ve developed a vision system that can recognize all this different material and then we work with our partners who provide the robots to get some hardware to actually do the separation,” Matanya Horowitz, founder of AMP Robotics in Denver, said.
The vision system, which is controlled by AI, does not operate through programming. Instead, it learns and as it goes and can distinguish between what is a carton material and what is not.
“It is able to learn logos, shapes, textures,” Horowitz explained. “It can actually learn things like Pepsi symbols are highly correlated with plastic or Horizon milk is very correlated with cartons.”
The pilot program started in late 2016 and, through fine-tuning and adjustments, has achieved a pickup rate of 60 cartons per minute, which is higher than the human average of 40 picks per minute.
Response to Nielsen data
Posted by Mary Ellen Shoup,
Source for Nielsen data
Posted by David Sudolsky,