Sort it out: Brand owners and researchers unite on AI tech for flexible packaging circular economy

31 Mar 2022 --- Nine major brand owners, including PepsiCo, an independent test and research center and two universities, are uniting on a project to assess how AI technology can help improve packaging sorting capabilities and support the circular economy.

The Perfect Sorting Consortium will test AI technology that could help separate packaging that is not currently sorted efficiently. Beyond the two-year project, the consortium’s ultimate goal is to make this AI decision model available for wide use in European sorting plants in Europe.

The project involves PepsiCo, Nestlé, Danone, Colgate-Palmolive, Ferrero, LVMH Recherche, Mars Incorporated, Procter & Gamble, Michelin, National Test Centre Circular Plastics (NTCP) (Netherlands), Ghent University (Belgium) and Radboud University (Netherlands).

PackagingInsights speaks to Gareth Callan, PepsiCo Europe’s sustainability packaging manager, to understand why improved waste sorting is critical to creating a flexible packaging circular economy and the challenges facing the consortium.

“AI is already addressing some of society’s biggest challenges and may bring the step-change needed to advance material sorting in Europe with the ultimate goal that more is recycled into higher value materials,” he says.Gareth Callan, PepsiCo Europe’s sustainability packaging manager.

Collaborative action
Stakeholder collaboration will be essential to scaling this technology to European sorting plants, stresses Callan. “Undertaking this work through pre-competitive collaboration is vital to ensure we crack the code on AI decision-making technology that works for all and can be rapidly scaled,” he explains.

“The infrastructure and sorting challenge can not be solved by one company alone. Many nascent technologies use AI/machine learning technologies to identify packaging. The challenge is to find a scalable AI model that is open source with a brand owner-supported packaging database. Any viable solution will require widespread support of sorting technology suppliers.”

The Perfect Sorting Consortium is organized to enable open pre-competitive collaboration to evaluate, develop, test and prove the optimum AI technology for scalability.

“The goal is to successfully test the technology in an industrial sorting plant. This solution could preferably be implemented in existing sorting technologies with minimal costs and complexity such that a swift entry into the market is possible,” adds Callan.

Decision time
Material sorting infrastructure in Europe is predominantly manual, meaning that some materials are not sorted and recycled, including food and non-food packaging. According to Eurostat, more than 79 million tons of packaging waste were generated in Europe in 2019 while 65% were recycled. Among all materials, plastic packaging waste represents more than 15 million tons, and the associated recycling rate was 41%.

The consortium aims to drive a change in infrastructure and sorting technology to ensure these materials can be identified, sorted and recycled by developing and testing an AI decision model. The project is not developing a sorting machine but a decision-making solution that can be implemented within an existing machine.PepsiCo Europe has pledged to eliminate virgin fossil-based plastic from its chip packs by 2030.

“The model involves testing and optimizing input measured by a range of sensors that detect different things: color, dimensions and shape, patterns, presence of metals, presence of different plastics or papers, both at the surface and below the surface,” explains Callan.

“The question is what permutations and combinations of sensor input lead to the most accurate identification. Brand owners are also sharing the specification of their packaging so that machine learning learns from accurate input. By training the AI to accurately identify a wide range of brand owner formats, the project will explore the limits of what machine learning can accurately detect and how this adds value to waste feedstocks.”

Product testing
NTCP will test a wide range of packaging products provided by the brand owners using their own flexible and modular sorting line with industrial equipment. These assessments will be used to further develop the packaging sorting model of the University of Ghent.

NTCP, together with the universities of Ghent and Radboud, will develop an AI decision model to detect, identify and classify packaging beyond the current sorting streams. The consortium members also aim to involve different technology providers and waste management companies during the project.

Current sorting techniques have limited performance as they are based on a few parameters, such as the type of material, rigid or flexible properties, and color. Also, some packages are not efficiently separated, which is the case for food versus non-food packaging or flexible multi-material packaging.

Disruptive sorting projects
Over the past few years, several projects focusing on developing disruptive technologies, such as digital watermarks, have been initiated to enhance packaging waste sorting in the EU.

The consortium aims to improve on manual material sorting with AI-based tech.The Perfect Sorting Consortium is part of PepsiCo’s ongoing work to collaborate with waste managers, re-sorters and recyclers to help build a circular economy in Europe. The project accompanies the HolyGrail 2.0 digital watermarks initiative to understand alternative routes and the best solutions for rapid scalability. The HolyGrail 2.0 project recently announced it has successfully validated its prototype detection unit from semi-industrial testing mimicking real-life conditions.

“Digitally watermarked packaging has the potential to deliver the same (or even more accurate) packaging identification outcomes for sortation,” continues Callan. “Watermarking also offers potential in smart-EPR and consumer engagement both from recycling, nutritional and promotional perspectives. Holy Grail remains an experimental proof-of-concept initiative that PepsiCo is fully committed to lead.”

“History tells us that not all advanced technologies are ultimately scalable to market, so Perfect Sorting offers the opportunity to fully evaluate alternate perspectives to finding solutions to the rapid and scalable technology required to unlock value in flexible packaging.”

“We do not know that the right answer might be at this point, so we are committed to evaluating both to explore to what extent they are complementary or not.”

In January, PepsiCo Europe pledged to eliminate virgin fossil-based plastic in all its chip packs by 2030. CEO Silviu Popovici called for “an appropriate regulatory landscape” to support flexible packaging circularity goals.

By Joshua Poole