TOMRA Recycling launched an AI-native platform and expanded its GAINnext ecosystem with deep learning tools to boost sorting accuracy, efficiency, and recycling profitability.
TOMRA Recycling has launched a next-generation AI-native platform from PolyPerception and expanded its GAINnext™ ecosystem with new deep learning applications, marking a significant evolution in recycling operations. These innovations, unveiled at IFAT 2026 and PRSE, represent a shift from traditional sensor-based sorting to intelligent, connected, and AI-driven facilities. The move coincides with TOMRA increasing its stake in PolyPerception to a 51% majority, aiming to create a direct link between real-time data and physical sorting actions.
The new AI-native platform, an advanced iteration of PolyPerception's Waste Analyzer, introduces a natural language interface that allows operators to interact with plant data conversationally. This eliminates technical barriers, enabling users to ask questions in plain language and receive immediate, context-aware answers with data breakdowns. Beyond merely reporting data, the platform possesses "writing capabilities," functioning as an active agent within the plant to create custom quality reports and set operational alerts based on its deep understanding of recycling processes. It also offers full transparency by integrating plant data into existing management systems, allowing managers to query waste statistics or purity levels through their own dashboards. Additionally, the platform features powerful search methods like "similarity search" to identify visually similar problematic objects, such as fire hazards like batteries, without requiring new AI model training, and text/brand search for specific items.
Complementing this, TOMRA is introducing new deep learning applications for its GAINnext™ ecosystem, designed to address long-standing industry bottlenecks where conventional sensor-based sorting has reached its limits. Key applications include advanced sorting for food-grade PET trays, where the system can distinguish between various tray types based on shape and use, achieving over 95% purity. This breakthrough transforms PET tray sorting into a viable business case for recyclers. In the metals sector, a high-precision application targets "copper meatballs," automatically identifying complex copper-steel composites even in challenging streams, thereby upgrading rebar-grade scrap to premium furnace feedstock and supporting steel decarbonization. A third application significantly improves used beverage can (UBC) aluminum recovery from packaging streams, offering up to 33 times more throughput than manual sorting and delivering 98% purity or higher.
These developments have substantial industry-specific and economic impacts. They redefine how recycling plants operate, boosting operational efficiency, real-time data analysis, and sorting precision. The enhanced accuracy and purity lead to higher-value recyclates and create new material streams, fostering greater material circularity in industries like aluminum. The automation and improved decision-making tools are expected to increase profitability for recycling facilities by optimizing processes and reducing operational costs.
