In a previous Advisor, we discussed the application of AI — in conjunction with computer vision, machine learning (ML), and robotics — for optimizing plastic recycling. This Advisor examines specific examples of how companies are utilizing these technologies to develop commercial solutions for solving the plastics recycling problem.

Intelligent Recycling 

Companies are applying AI, computer vision, ML, and robotics to develop intelligent recycling platforms designed to enhance the efficiency and profitability of the recycling process by:

• Improving sorting accuracy

• Increasing process efficiency

• Optimizing the recovery of more valuable materials

• Enhancing processing (i.e., the ability to process/recycle more complex types of plastic materials)

• Facilitating data-driven decision-making

In practice, intelligent recycling platforms employ advanced cameras placed above conveyor belts and ML algorithms trained to identify plastic objects (and other materials) by shape, color, and labels as they move through the various stages of the recycling process. They are typically paired with robotic sorters in the form of mechanical arms and other machinery, which serve to physically sort through the various types of plastic objects (often among other waste) by picking them off the conveyor belt and placing them into bins earmarked for the appropriate form of recycling.

By automating the sorting process, AI-powered intelligent recycling platforms can significantly reduce the time required for sorting plastic waste compared to manual sorting. This helps optimize throughput, offering greater operational efficiency in recycling centers. Moreover, unlike manual sorting, which is often inconsistent and error-prone, they offer a consistent and reliable means for identifying and separating materials. This is key because consistent and accurate identification and categorization helps ensure a higher quality of recycled product.

AMP Robotics

AMP Robotics offers hardware and software solutions utilizing AI-based image analysis to sort recyclables with high accuracy and recovery rates. AMP’s technology is modular in design to make it more practical for recyclers to integrate AI-based sorting capabilities into their existing facilities.

AMP has developed AI algorithms to enhance the functionality and increase the recovery performance and reliability of robotic sorting systems. These “robotics detection models” are designed to teach robots to detect plastic waste in the waste stream based on various attributes, including color, shape, material, and brands, increasing the sorting rate significantly.

AMP also develops advanced ML targeting algorithms to determine the optimal grip area for each item identified (in the waste stream) by a robotic sorter, based on the object’s material features and condition. The ability to target and guide a robot to the desired grip area is important because it helps increase yield by enabling the robot to learn to avoid creases, holes, tears, crushed areas, and other difficult-to-grasp locations on recyclable objects like plastic bottles and packaging. The robotics detection algorithms and advanced targeting algorithms both have the ability to learn from experience and adapt to new robotic gripping technologies, which are continually evolving to meet recycling needs.

AMP also offers analytics and a Web portal for analyzing data generated by its computer vision and robotics systems. These tools allow recyclers to gain insight into their material sorting and robotic operations. This includes analyzing real-time material flow throughout the key stages of the sorting process and for robot performance measurement and optimization.

Glacier

Glacier is developing computer vision and robotics technology to automate the sorting of recyclables and to collect real-time data on recycling streams for recycling companies and consumer brands.

Key to this effort is a proprietary AI model and computer vision technology capable of identifying more than 30 categories of recyclable materials in the waste stream. These range from broader categories like cardboard and plastic containers to more specific items like cat food cans and toothpaste tubes.

Glacier’s AI model works with a custom robotic sorter, which can also sort a range of items, including soft plastic materials such as grocery bags and trash bags. This is particularly significant because such soft plastics have proved difficult to sort using conventional automated recycling technology due to their tendency to get caught in mechanical waste sorting and separation machines, causing equipment breakdowns and contaminating other recyclables.

In March 2024, Glacier raised US $7.7 million in funding, including from Amazon’s Climate Pledge Fund. Amazon and Glacier have also partnered to develop and pilot technology in an effort to increase the use of recycled materials in packaging. Glacier’s technology is designed for deployment within existing recycling facilities, with the company’s robot taking up about the same footprint as a person.

Greyparrot

Greyparrot is a start-up focused on developing AI-powered analytics to optimize waste sorting processes and facilitate data-driven facilities management. Its waste analytics technology captures real-time images and outputs AI-generated data pertaining to waste streams. In effect, it provides visibility into the composition of waste processed within a facility; recyclers can identify and classify the value of the waste streams based on the types of materials, brand, size, mass, function, monetary value, and emissions potential of each piece of trash in the waste stream.

An automated alerting capability serves to notify operators of declines in purity and other sudden changes in waste material composition via a Web portal, SMS, or email alerts. By identifying changes in material composition against set thresholds, the system can alert operators to quality issues and potential blockages. This allows facility operators to act proactively in order to prevent waste stream sorting downtime.

Additional analytics analyze data generated from recycling operations for facilities management. Such guided decision-making lets operators monitor throughput rates, product purity, and the loss of valuable material due to residue on materials.

Recycleye

Start-up Recycleye is using AI to enhance the efficiency of the waste sorting process by enabling the automated recovery of more materials — including complex plastics that previously have proved difficult to identify and sort using traditional methods.

Of particular interest are the company’s machine vision and robotics systems designed to identify food-grade and non-food-grade plastic packaging in the recycling stream. This is significant. Before the use of AI, it had been very difficult (if not impossible) for mechanical sorters to screen and sort food-grade and non-food-grade packaging with an acceptable degree of accuracy. Consequently, a great deal of such packaging waste tended ended up in landfills. The ability to do so improves the quality and value of recycled materials, which helps increase the circular recycling of these products.

Recycleye has partnered with robotics manufacturer FANUC to develop robotic waste recycling equipment powered by its computer vision sorting technology (Recycleye Vision). This includes the Recycleye QualiBot designed to sort dry mixed recyclables consisting of plastics, non-ferrous, and fiber materials.

Conclusion

The companies examined in this Advisor by no means represent the only ones applying AI, machine vision, and robotics for recycling plastic (and other) waste. However, I do believe they provide a good overview of how companies are implementing these technologies within commercial solutions intended to optimize and enhance plastics recycling operations.

The bottom line is that AI, in consort with machine vision and robotics, is proving crucial for making high-volume and high-accuracy plastic recycling practical in production environments. This includes enabling facilities to recycle forms of plastic that were previously too difficult or impractical to sort using conventional automated recycling methods. As the volume and types of plastics produced by industry continue to grow, AI, machine vision, and robotics will become indispensable for dealing with the plastic recycling problem.

Finally, I’d like to get your opinion on the use of computer vision, AI, and robotics for plastic recycling and recycling in general. As always, your comments will be held in strict confidence. You can email me at experts@cutter.com or call +1 510 356 7299 with your comments.