How Machine Learning and Robotics are Solving the Plastic Sorting Crisis

By Carrissa Pahl Published on Aug. 31, 2020

“If present trends continue, by 2050, there will be 12 billion metric tons of plastic in landfills. That amount is 35,000 times as heavy as the Empire State Building.” If that isn’t a rallying cry, what is?


The United States makes up 4% of the world’s population but produces more waste-to-population than any other country at 12%. “ India and China make up more than 36% of the world’s population and generate 27% of the world’s waste. To make things worse, the U.S. needs help when it comes to disposing and reusing waste. Each year, approximately 90 million tons of extremely useful recyclable products are thrown away and sent to landfills,” said Chris Wirth, VP of Marketing for AMP Robotics.


China Rejects U.S. Recycling

Last year, China stopped accepting the majority of U.S. recycled plastic products. Why? Contamination problems. Contamination is a huge issue for recycling plants. For example, if you were to put food scraps in a recycling bin or any material that was not recyclable for that matter, you would be contaminating any high-quality plastic that could otherwise be recycled and reused. Less than 8.4% of plastics are recycled. Why is that? 

Less than 8.4% of plastics are recycled.
  • Recycling as an industry has faced tremendous challenges when it comes to commodity and re-selling markets.

  • In the past few years, the problem of recycled materials being returned to the U.S. from other countries due to contamination and bad quality feedstock has shown clear discrepancies in the quality of what is recycled here in the U.S.

  • The cost to recycle is expensive so most people figure it’s easier to dispose of waste in landfills instead.

  • Billions of dollars are lost to landfill because of contamination of products and poor recycling techniques.

Historically, China recycled the bulk of U.S. waste products. By rejecting U.S. waste, this has forced a sudden need to solve the recycling crisis. Instead of properly separating recyclables, the U.S. dumps all waste into a single bin, impartial to sorting.  As a result of China rejecting U.S. materials, many communities have given up on recycling all together due to rising costs. 

This is the problem that companies are trying to solve. As companies try and tackle this issue, many of them are turning to AI and machine learning to come up with solutions. 


Is Robotic Sorting the Answer?

Robotic recycle sorting uses artificial intelligence and robotics to sort plastics so humans don’t have to. With advanced cameras and technology, these companies are counting on robots to sort recycling, as well as reduce any health risks that come along with human labor. According to a report at the University of Illinois School of Public Health, recycling workers are twice as likely to be injured on the job as other workers. 


How does robotic recycling work?

Cameras and high-tech computer systems that are trained to recognize specific objects will guide robots’ arms over conveyor belts to reach their target. Oversized fingers with sensors that are attached to the arms are able to snag cans, glass, plastic containers, and any other recyclable items out of the otherwise garbage and place them in their respective bins. 

Recycling robots are still assisting humans, but companies have found that they can work two times as fast as humans. Industry leaders have developed robots that can identify different colors, textures, shapes, and sizes of plastic materials and make it easier to sort waste. 

This technique can increase the quality of material output and in some cases double the resale value. As quality standards get stricter, companies are working fast to find reliable solutions. One city in particular is trying to beat the curve. San Francisco is racing to become the first U.S. city to reach zero waste. Recology, a company that runs a large plant on the San Francisco Bay, just completed an $11 million upgrade and plans to invest another $3 million this year in high tech optical sorter robotics.


Pros and Cons to Robotic Sorting


Some advantages of robotic recycling include:

  • Reduced reliance on manual sorters: Materials recovery facilities across the U.S. struggle to hire and retain workers. An AI-powered robot can replace the 1.5 workers that are needed to pull cartons off the line. Those workers can then be moved to other places that need attention, thus speeding up the process.

  • Quicker sorting off the line: Current robotic techniques include using a camera to look at each product coming off the line and using data that has been stored over time to analyze where it will go. This also means that the robots can constantly update and add to their data of materials, which will help down the line. The benefits of AI are rapidly increasing and soon, robots may be able to make all the necessary adjustments on their own.

  • Improved knowledge: Optical sorters operate by detecting specific types of material. This will increase knowledge over time on what kind of materials are coming in each day and how they differ from each other. There is also the potential to inform other plants across the country who are also using robotic sorting methods. 

  • Advantages of AI: These robots will be able to store data and process it much faster than humanly possible. This will add to a database that can then be used by the greater recycling community in order to better their output.

  • Quality control: Using robots can ensure quicker, more precise, and better quality of plastics being picked out and sent to the correct bins. This will also ensure that more plastic can be reused. Having quicker and more precise “hands” will provide more opportunity for output. 


Every good comes with some bad. Some disadvantages of robotic recycling include:

  • The cost: Right now, facilities must process significant amounts of material just to justify the cost of these robots. However, over time, these robots will be able to lower hiring costs and the output will outweigh the costs

  • Supply and demand: Buyers often won’t commit to taking on a recovered material type until they know they can rely on sufficient tonnages being provided by processors. On the other hand, processors don’t want to devote man hours to a new stream until they know they can make a profit.

  • Robots need a constant power source: This may add to the already expensive cost and shy people away from incorporating them into their sorting plants. 


How Does Machine Learning Play a Role?

Machine learning is an aspect of artificial intelligence. It is defined by Expertsystems as an application of AI that provides systems the ability to automatically learn and improve from experience without being explicitly programmed. Machine learning focuses on the development of computer programs that can access data and use it to learn for themselves. 


This allows for quick adaptation and effortless updates to the technology. Machine learning also minimizes human interaction. This leaves more opportunity for those workers to focus on more important issues or engage in higher-paying, more rewarding jobs. Machine learning has become extremely popular in many areas of work including  industries such as government agencies, health care, retail, transportation, and of course the recycling industry. 


Waste Sorting Startups 

How AI and Robotics are Solving the Plastic Sorting Crisis - AMP Robotics

At AMP, technology is all about productivity. They design solutions so companies can overcome limitations of manual processes and help maximise operations. Driven by innovation, AMP is constantly seeking ways to improve automation with the latest applications of advanced AI. AMP has three different systems in use. 

  1. AMP Cortex™

    1. Cortex™ is a high-speed, intelligent robotics system designed to meet the demanding needs of today’s recycling operations. Guided by AI, its robots intelligently perform physical tasks of sorting, picking, and placing material.

  2. AMP Neuron™

    1. Cortex is powered by the AMP Neuron™ artificial intelligence platform. Neuron combines machine vision with deep learning to capture and recognize the unique characteristics of objects within a mixed material stream.

  3. AMP Insights™

    1. Data captured by Neuron is made available via AMP Insights, an online data visualization tool. 4Use it to monitor activity, measure performance, and help you make informed decisions about your operations.


How AI and Robotics are Solving the Plastic Sorting Crisis - cleanrobotics

CleanRobotics is focusing on how to create a greener tomorrow. They are creating sustainable, innovative, tech driven solutions to persistent environmental problems. CleanRobotics has created a product called TrashBot. TrashBot is an autonomous system that uses robotics, computer vision and artificial intelligence to detect and separate landfill from recyclables. It does this more accurately than human beings, captures high quality waste data and it lets staff know when it’s getting full. Cloud connectivity allows individual units to learn from the global TrashBot fleet, becoming more intelligent over time. It also has a monitor for corporate communications, education, and advertising.


How AI and Robotics are Solving the Plastic Sorting Crisis - zenrobotics

A global leader in smart robotic recycling, ZenRobotics was the first company to apply AI-based sorting robots to a complex waste-sorting environment. Their robots are powered by their own AI software and make recycling more efficient, accurate and profitable. ZenRobotics has two products: the Heavy Picker and the Fast Picker. The Heavy Picker is their strongest robot. It can pick up to 6,000 pieces of waste in an hour. The Fast Picker can pick up to 4,000 pieces of material per hour and is beneficial for maximizing material recovery.  


How AI and Robotics are Solving the Plastic Sorting Crisis - greyparrot

Greyparrot has developed a waste recognition software to monitor, audit, and sort waste at scale. Using AI-powered computer vision software, Greyparrot automates the measurement of waste streams. Real-time analysis on 100% of waste flows is helping waste managers save cost, increase revenue, and mitigate against risks. Their goal is to empower all stakeholders in the waste management process with actionable insights to increase recycling and recovery rates. Greyparrot’s end-to-end solution is deployed on moving conveyor belts in sorting and recovery facilities. They also provide waste analytics to automate the current manual auditing process and provide new insights previously unavailable to waste managers, producers, and regulators. 


everest labs ai - machine learning and robotics in plastic sorting

Combining their knowledge in recycling, computer vision, and AI, Everest Labs works to provide cost effective and space efficient robotic and business intelligence solutions. Their goal is to advance recycling techniques while helping the planet and contributing to the end goal of ending the plastic waste crisis. Everest Labs uses in-house recycling expertise, deep data, and evolutionary machine learning to provide business intelligence and automation. They strive to be reliable and economical. 


If you want to learn more about our end plastic waste initiative in partnership with the Alliance to End Plastic Waste, check out our website.