AI Helping Recyling Industry Improve Accuracy, Speed Sorting Rate

TOMRA got its start in 1972 as a provider of reverse vending, aka bottle collection, machines.

By AI Trends Staff

AI is helping the recyling industry improve the accuracy of identifying specific types of plastics and other materials, including items contaminated with food and other substances, as well as speed up the sorting rate.

Smart robots, sensors and vision systems fortified with machine learning software are creeping into production at recycling facilities in Colorado, Japan and Europe.

Here are two companies talking up the potential to make the act of processing everything from plastic to demolished construction materials far more efficient and scalable, according to an account in GreenBiz: five-year-old startup AMP Robotics, a machine learning and computer vision specialist headquartered in Louisville, Colorado; and a Norwegian company, TOMRA, which got its start managing reverse vending machines that uses sensors to endow its food sorting and recycling systems with more intelligence.

A New Vision for Sorting

As its name suggests, AMP Robotics’ innovations lie in how it’s rethinking recycling robots. Founder and CEO Matanya Horowitz began receiving grants back in 2014 to research and develop vision systems that could improve the accuracy of separating items with machines rather than by humans. The company’s equipment is “trained” by being shown millions of images — everything from logos to box shapes to dyed plastics.

“If you can teach a person to distinguish something, you can teach our vision system to distinguish it,” said Horowitz to GreenBiz.

The idea, he said, is to help facilities become far more specific about separating streams of waste, which could allow operations to capture revenue from entirely new sorts of services.

For example, the technology — using a combination of light and machine learning software — could be used to sort out colored whipped cream tubs or yogurt containers from clear plastics. It can even identify items that carry a specific brand logo. One early adopter, Alpine Recyling in Colorado, recently was able to add coffee cups to the mix of stuff that its facility can handle. These levels of specificity could be valuable for consumer products companies seeking either to put their own product packaging back into circulation or to buy specific types of plastics.

“We can track what is truly being recycled,” Horowitz added, and that could help provide insight into where better collection systems — and messaging — might be needed.

AMP’s latest technology is a dual-robot system called Cortex, which the 35-person company will sell for municipal solid waste, electronic waste and construction and demolition applications. The equipment can sort, pick and place items at a speed of 160 pieces per minute. More important, it will allow facilities to tackle a process that typically has been very difficult to scale — separating post-consumer fiber from cardboard to sheets of paper.

Horowitz is cagey about how much money his company has raised, although its backers include Closed Loop Partners, and he called out the Alphabet company Sidewalk Labs during our conversation. Likewise, he won’t talk about the cost of his firm’s technology, pointing out that customers are seeing a payback of less than two years and that it sorts at the rate of two people.

That latter statistic might give pause to those concerned about the job-elimination potential of robotics technology, but Horowitz says recycling facilities often have high rates of turnover. “Many facilities are run underutilized,” he said.

AMP Robotics is also touting applications in the construction sector. Earlier this year, it disclosed a partnership with Japanese waste management company Ryohshin to sell AI-driven robots for recovering materials out of demolition debris — including wood, metal, electronics and concrete.

TOMRA Seeks Help from AI to “Create Value Out of Waste”

TOMRA, which recently joined the Alliance to End Plastic Waste, is credited as the inventor of near-infrared sensors for sorting applications. It holds close to 80 patents, and has extensive experience with sensors that can provide information about moisture or chlorine levels as items are being washed and sorted. The company’s technology is installed in roughly 100,000 locations worldwide; aside from recycling and collection, the company sells to food operations that need to sort things such as fresh produce.

“We urgently need to transform the recycling industry by creating value out of waste,” said Stefan Ranstrand, TOMRA president and CEO, in a statement. “In some markets, recycling rates are as high as 98 percent, and some consumer goods companies are now making new products out of 100 percent recycled materials. But this is only a very small part of the picture, and much more must be done to preserve our world for generations to come.”

Two of TOMRA’s most recent innovations include a material recognition sensor that can sort single-layer polyethylene terephthalate (PET) trays (think cafeteria trays) and a new laser feature. The latter development helps TOMRA’s systems detect black objects that are typically hard to ID, as well as those with certain shapes, such as silicon cartridges. Over time, the AI associated with these systems will be able to detect the presence of films within rigid plastics, according to the company’s website.

“If you really need to sort out materials to the highest quality possible, you need to be able to do this,” said Volker Rehrmann, executive vice president of TOMRA and head of the 4,000-person company’s newly formed circular economy team.

AI has been part of TOMRA’s technology for some time — some level of algorithmic detection was introduced as part of its original bottle collection machines. The company collects 40 billion used beverage containers annually through those reverse vending machines. So it has plenty of legacy experience, which will be vitally important for finding value in the wide range of plastics that now exist in the world.  

“Without going into the details, it now needs to be much more accurate than it used to be,” Rehrmann said.

That’s because even though recycling is big business — an estimated $110 billion in America alone last year — he estimates that only 2 percent of those materials are valued as a resource, not waste, within a closed loop economy. That’s a problem that the Alliance to End Plastic Waste is determined to address. “They have realized that change is necessary,” Rehrmann said.

History of AI in Reycling Sorting

AI’s use in sorting started with systems in the 1970s to 1980s, according to an account in recyclingtoday. These systems were based on optical sensors and electronics that compared gray values or colors. Based on the ratio between these colors, the electronic circuit would make a rules-based decision whether to keep or eject the material. For instance, the first reverse vending machine (RVM) recognized the shape of a bottle based on the shadow it generated, which was detected by prepositioned optical sensors.

In the early 1990s, pixel-based classification of gray-scale and color camera images was used in combination with custom-made electronics, which limited the capabilities of the AI in terms of thresholds and decisions. With the emergence of personal computers (PCs), it became possible to use this technology for the classification of images.

Customized camera technology was used to acquire specific spectral properties and better clustering possibilities, leading to improved AI accuracy. This made it possible to assign each pixel with a specific class of material based on its spectral content. Color was no longer the only identifying criteria.

This technology was then combined with object recognition in the late 1990s, which made it possible to cluster different pixels of similar properties and combine them into an object.

By the 2000s, hyperspectral imaging systems became available, and the power of PCs further increased. Artificial neural networks (ANN) started to become available for classification problems in data processing. Based on previously trained samples for the specific application and machine, this class of AI could now combine different features and properties to make one classification. As a result, more complex materials could be detected and another level of sorting accuracy was achieved.

Later in the 2000s, so-called support vector machines (SVMs) became available. Although it sounds like a physical machine, these are mathematical models that allow a machine to define clusters in multidimensional space. Storing the results in tables on the physical sorter improved the performance again.

Common to all previously mentioned forms of AI used for sorting is the fact that the so-called training or learning aspects of AI must be supervised. In the very simple example of the RVM from the 1970s, the engineer had to physically place the optical sensor in the correct location, and a set of labeled samples needed to be available for teaching the system before putting the sorter into operation.

How AI Is Employed in Recyling Today

Today, the initial teaching of the system requires a computer vision engineer to define the relevant features for the sorting task. This generates feature vectors from the image data, which are then used in conjunction with the labels for automatically training the ANN or SVM. Because the training is done automatically without interaction by the engineer, this approach is called unsupervised learning.

The next step in the evolution of AI in sorting is deploying deep learning methodologies that became available in the 2010s and are now used in a range of applications. These types of networks were invented decades ago. Because of a massive increase in processing power in modern graphics processing units and millions of generally available and labeled images, it is now possible to apply them to practical problems.

So-called deep convolutional neural networks are still an ANN; but, compared with the early derivations, they have a much larger number of layers and neurons. Consequently, the networks are more powerful. However, they also require a lot more training data than traditional approaches.

The major advantage of convolutional neural networks is that the feature extraction step also is performed automatically during the training of the network. As a result, a computer vision engineer is no longer required to manually define the features relevant for the task. Typically, the first layers of the network generate features, which are integrated into more complex features in the following layers and then classified in the last layers.

These networks can be combined almost like building blocks, with each one being pretrained for a certain task. By doing this, the design can be adapted to the application at hand. Deep learning technologies have a major impact in image recognition in particular.

Where AI Is Taking the Recyling Industry

The current phase of AI development—especially the deep learning aspects—will enable the recycling industry to tackle currently unresolved challenges.

Today, a handpicking station at the end of a line is still needed to improve the quality of the final product to the desired level. An example could be seen with silicon cartridges, which are not desirable in a polyethylene stream. To pick them up with a robot, or to eject them through a last optical sorter, they would need to be detected first.

For this capability, AI and deep learning will play an important role in improving efficiency. Combining these new forms of AI with the potential of big data (e.g., with the data we already can collect from the machines today) will open even more opportunities to increase production, reduce cost and improve quality.

There’s this idea that today’s AI is like finding a free lunch and an ugly duckling. Both ideas are actual mathematical theorems that relate to the topic of artificial intelligence.

The first theorem basically states that no single AI solution is superior to all the others for a specific application. Each solution can have certain benefits that come at the cost of some disadvantages elsewhere—hence, there is no such thing as a “free lunch.”

The ugly duckling theorem is similar, stating that no optimal feature set exists for all applications. Even if we could find generic AI that solves many different challenges, it would not fit at least one application or problem and would not provide a suitable solution—making it the “ugly duckling.”

With this in mind, we should stay modest in our expectations regarding what is possible with convolutional neural networks and deep learning. Plenty of examples are available where deep learning is solving difficult, loosely structured recognition problems, but with other sorting tasks other AI approaches will have better performance. Finding the right combination of different types of AI was key in the past and will remain a key to ensuring the best sorting performance for recyclables in the future.

Artificial intelligence has been deployed in the recycling industry for quite some time. Yet, the possibilities that deep learning can offer when the fields of machine vision and machine learning are brought under one umbrella are new and exciting for the industry.

(Ed. Note: The authors  of the recylingtoday piece are vice presidents with research responsibilities for Germany-based TOMRA Sorting GmbH,, part of Norway-based TOMRA Systems ASA.)

Read the source posts in GreenBiz and in recyclingtoday.