Sonoco, a global packaging solutions leader, partnered with AMP Robotics to create a new material category within AMP’s neural network specific to rigid paper containers. 

Sonoco is a global packaging solutions leader headquartered in Hartsville, South Carolina. The company provides diversified consumer and healthcare packaging, industrial products, protective packaging, and packaging supply chain services and is the world's largest producer of rigid paper cans, tubes, and cores. Together, Sonoco and AMP created a new material category within AMP’s neural network specific to rigid paper containers. 

Impetus: Enhance recovery for unique packaging format 

Thanks to its long history in the packaging industry, Sonoco recognized there’s a great deal of material that’s misunderstood, or not collected due to a lack of understanding. The company has taken a deliberate, focused approach to increase recovery of its packaging. This approach led Sonoco to partner with AMP Robotics to help improve collection and sortation of a particular packaging type that’s highly recyclable: rigid paper containers. These are items like coffee, nut, and snack containers; refrigerated dough; and powdered infant formula.  

Sonoco sought a solution for the sortation of rigid paper containers applicable to a variety of MRF configurations. It was important to identify a technology solution that materials recovery facilities (MRFs) could drop into their existing operations.

Solution: AI technology that continuously learns and improves

AMP’s proprietary AI technology, AMP Neuron, works by perceiving images of recyclable materials on conveyor belts within recycling facilities. Looking for specific colors, shapes, textures, logos, and more, the system recognizes patterns correlated with material type. AMP’s neural networks learn to identify objects in the same way a human does.

Training a neural network to detect objects in the recycling stream is not easy—but it’s an entirely different challenge when you consider the physical deformations that these objects can undergo by the time they reach a recycling facility. They can be folded, torn, smashed, or partially obscured by other objects.

AMP’S data annotation team analyzes the images it collects and categorizes the objects belonging to each material type. The company continuously collects samples from all the facilities that use AMP’s robotics systems, and then annotates them, adds them to its database, and retrains its neural networks. AMP also tests its networks to find models that perform best on target material and conducts targeted additional training on materials that its systems have trouble identifying correctly.

Process: Collaboration, training to develop custom label

Over the years, Sonoco optimized the design of its rigid paper cans to provide needed product protection and shelf-life while minimizing environmental impact. The use of recycled steel has a 45% lower environmental impact than producing the equivalent amount from virgin material. Additionally, when compared with landfilling, recycling the paper container with steel or paper bottom through the appropriate stream has a greater than 40% lower environmental impact than landfilling. The can body of the rigid paper container comprises 100% recycled paperboard, of which 85-90% is sourced from post-consumer recycled content. The robust structure of the rigid paper container allows for use of recycled materials while also maintaining product freshness and protection through distribution, and is ultimately recyclable through steel or paper streams. Working with AMP to develop a new material category was the next step in ensuring the material could be recovered in MRFs. 

The process began with a discussion of goals between Sonoco’s sustainability team and AMP’s AI and data team. Early on, Sonoco made sure to provide AMP with plentiful images of the containers so AMP could train its neural network to identify the material and conduct early trials. That step progressed to an interactive series of trials in MRF environments. Through this process, Sonoco learned an effort like this requires revisions and many rounds of tweaking—and even iterations that aren’t perfect. For example, the training process may reveal that the robots identify BBQ snack cans quite accurately, but need more work on other varieties. Yet all along, the robots are picking more of this material type than was recovered before. 

In addition to being a manufacturer, Sonoco is a recycler, operating more than 40 recycling facilities globally and collecting nearly three million tons of paper, plastic, metal, and other materials annually. The company has installed an AMP Cortex™ intelligent robotics system in one of its facilities, with more deployments planned; this recycling relationship enabled Sonoco to directly test and leverage AMP’s technology. 

Results: More post-consumer recycled content, pathway for growth and performance improvements  

AMP’s AI platform has recognized more than 50 billion containers and packaging types in real-world conditions, and it’s constantly learning and improving. The implications of AI in the sorting process are profound, and AMP is working with companies across the spectrum of recyclable materials to adapt its AI platform to new form factors of materials, or the specificity of a manufacturer or brand, and disperse the sorting capability to MRFs with its robotic sorting systems. Any MRF with an AMP Cortex™ system can accurately and efficiently sort Sonoco’s paper can to the desired stream.  

Moreover, in addition to lowering the environmental impact associated with resource extraction to create new packaging, Sonoco’s collaboration with AMP shows how manufacturers are directly influencing what’s recoverable and taking advantage of the ability to capture more of their specific packaging so they can incorporate the recycled content into new packaging. The Sustainable Packaging Coalition recognized the partnership between AMP and Sonoco with a 2022 Innovator Award for Innovation in Recovery.