You might want to rethink that congratulatory pat-on-the-back you give yourself when choosing to recycle instead of trash your venti Starbucks cup.
Recent studies have found that up to 79 percent of plastic waste created in the U.S. goes to landfills instead of being recycled.
Yes, even the plastic that makes it to recycling centers is not always recycled.
This shocking discrepancy stems in part from the tedious and tiring labor that is typically required to manually sort tons of recycling that centers receive daily. During the pandemic, this volume increased while the availability of workers decreased, due to sickness or safety concerns.
This is a problem that AMP Robotics, a company founded in 2014 by CalTech Ph.D. Matanya Horowitz, plans to solve. Using computer vision and machine learning, AMP Robotics has designed an intelligent three-handed robot that can analyze and sort 80 items of recycling per minute — a rate that is twice as fast as human sorters. The company recently closed a $55 million funding round, bringing its total funding close to $75 million.
Why it matters — By speeding up the process of sorting recyclable materials, AMP Robotics hopes to increase the total amount of materials that can be effectively recycled and reused, in turn, savings tons of emissions that these otherwise missed materials would accrue in landfills.
Ultimately, this technology could play a key role in closing the loop of recyclable manufacturing and creating a world free from waste (or, at least, recyclable waste.)
Here's the background — Ever since he was a child watching reruns of The Jetsons, Horowitz has been fascinated by robots and their ability to wield human-like intelligence, the founder says in a promo video on the company's website.
This fascination, combined with a concern over the environmental and economic state of recycling, is what led to the creation of AMP Robotics in 2014.
At traditional sorting facilities, workers arrive daily to hand-pick quickly moving recycled materials — whether that be plastic milk jugs, soiled pizza boxes, batteries, or even wood — from conveyor belts.
It's a grueling job that is naturally prone to error and omissions. The ever increasing volume of these materials as well as more strict purity specifications from foreign importers like China, meant that these facilities struggled to meet demands.
And as a result, omitted or impure materials are cast aside in landfills.
"There's a tremendous amount of value captured in paper, and plastic, and metal, that right now is lost at the landfill" explains Horowitz in a video. "The trouble is that the value of this material is really eroded by the cost of sorting it out in these recycling centers."
The big idea — When Horowitz started looking into these issues, he immediately saw an opportunity for a robotics solution.
Using computer vision and machine learning, AMP Robotics set-out to design a robot that could quickly recognize and sort any number of recycled materials and accurately send them to the correct sorting bin for processing quicker than a human sorter.
"Globally, more than $200 billion worth of recyclable materials goes unrecovered annually," Horowitz tells Inverse. "A.I.-driven automation enables the efficient recovery of more material, which increases recycling rates and reduces human impact on the environment."
"We’re developing new vision system offerings, innovating data capture for waste characterization to support operators, and expanding data access to packaging producers to help them achieve recycled content goals. A.I. opens up opportunities that make recycling a more important part of the waste ecosystem."
How it works — When it comes actually sorting these waste materials, AMP Robotics' technology has two main parts:
- AMP Neuron — the machine's brain and eyes
- AMP Cortex — the machine's body and hands
Using computer vision, the same kind of technology employed by self-driving cars to "see" the road, the AMP Neuron first "looks at" incoming recycling and using machine learning to quickly analyze what type of material it is and makes a decision on what it should do next — either call on its suction cup arm to scoop it up or let it move past.
Using reinforcement learning, each time AMP Neuron correctly sorts a piece of recycling it learns how to better do its job in the future too. This includes not only correctly moving items for sorting but recognizing items if they're squished or in poor lighting.
In addition to simply sorting the materials, AMP Neuron also keeps a digitized list of all items it saw in its cloud-based brain for managers to review afterward.
As for installing, the whole system is built on a metal frame that can be easily slotted over existing conveyor belts in just 48-hours, or the course of a weekend says the company. For $6000 a month, they claim that this technology can reduce costs for recycling centers by 70 percent.
As of April 2020, AMP Robotics reports that installations of its technology in more than 20 states have led to a reduction of half a million tons of greenhouse gases.
The Inverse Analysis — The road to shaking free from the shackles of waste is still far in the future, but technology like this could play a major role in improving how we manufacture and use these recyclable materials, and ultimately what their impact on our planet is.
With this green vision in mind, it will be important to remember the human cost of innovations like this as well and ensure that jobs created or shifted in these industries are not simply for roboticists and technicians but accessible to those who originally sorted these cups and bottles as well.