Small and medium-sized recycling facilities across Europe face a persistent challenge: automated plastic sorting systems exist, but they remain financially and technically out of reach. These systems require substantial capital investment, specialized programming expertise, and ongoing technical support that most smaller facilities simply cannot afford. The result is a two-tier recycling industry where large industrial operations benefit from efficient automation while smaller facilities rely on manual sorting with limited capacity and accuracy.
This technological divide represents more than just a business disadvantage for small recycling companies. It creates a bottleneck in Europe’s circular economy transition, leaving significant recycling capacity untapped precisely where it is most needed. The INCIRCULAR-funded CLARA project set out to change this reality by asking a fundamental question: what if automated plastic sorting could be configured through conversation rather than code?
Translating conversations into robotic coordinates
CLARA (Classification & Advanced Recognition Automation) is an AI-orchestrated vision system that eliminates the programming expertise barrier preventing small facilities from adopting automation. The system uses computer vision algorithms to detect and classify plastic waste on conveyor belts, but its innovation lies in how facility operators interact with it.
Instead of requiring technicians to write code or configure complex technical parameters, CLARA enables operators to define sorting workflows through natural language instructions. An operator can simply describe what materials need sorting and where they should go, and the system autonomously translates these conversational instructions into the precise technical configurations and robotic coordinates needed for automated sorting operations.
This transformation is powered by three specialized AI language models working together: one interprets the operator’s high-level sorting goals, another identifies which material properties are relevant for discrimination, and a third extracts structured sorting assignments from conversational descriptions. The system handles the technical complexity invisibly, presenting operators with an intuitive interface designed for practical facility operation rather than technical configuration.
From research to industrial reality
Development proceeded through intensive investigation of state-of-the-art technologies to identify the most suitable approaches for real-world SME deployment. The team evaluated cutting-edge Vision Transformer architectures and advanced depth estimation models, conducting comprehensive testing to assess their viability for industrial plastic sorting.
Some advanced models performed impressively in laboratory conditions but exhibited classification instability and processing delays incompatible with industrial throughput requirements. Our breakthrough came from combining proven computer vision algorithms with AI orchestration in ways that prioritized reliability and accessibility over technical sophistication.”
The final system architecture balances multiple considerations: it operates efficiently on cost-effective computing hardware suitable for SME budgets, maintains the processing speed necessary for industrial throughput, and delivers consistent classification performance across diverse operational conditions. The complete user interface was designed specifically for facility operators without technical backgrounds, providing visual workflow builders and real-time monitoring dashboards that make sophisticated automation comprehensible and controllable.
Real-world impact for small facilities
The system classifies plastics by type, detects material properties like color and rigidity, and tracks items temporally as they move through the sorting process. It integrates with robotic manipulation equipment to create a complete automation chain from waste detection through physical material handling.
Validation testing confirmed the system meets industrial performance requirements while operating on hardware within typical smaller businesses capital constraints. Perhaps more significantly, the natural language configuration approach achieved what traditional systems could not: making automated sorting operationally accessible to facilities that previously had no path to automation adoption.
For small recycling facilities, this represents a fundamental shift in available options. Automation transitions from an aspirational capability requiring substantial investment and ongoing technical support to a practical operational tool that facility staff can configure and manage directly. This accessibility has implications beyond individual facility efficiency, potentially unlocking recycling capacity across the European SME sector that currently remains constrained by manual sorting limitations.
The value of structured support
The INCIRCULAR funding enabled validation at operational scale that would have been financially prohibitive for an SME pursuing development independently. Testing automated sorting technology requires extended operational periods with real waste
streams and industrial equipment, representing investments well beyond typical research and development budgets for smaller companies.
Beyond financial support, participation provided access to the European recycling sector network, establishing connections with potential partners and early customers while building market understanding that would have been difficult to develop in isolation. The structured project framework with regular monitoring ensured development remained focused on practical deployment challenges rather than purely technical achievement.
The association with the INCIRCULAR initiative also enhances credibility when engaging with potential customers and partners, positioning the solution within a recognized European innovation framework rather than appearing as an isolated development effort from a smaller company.
Looking forward
The project delivered both a validated technical system and a structured commercialization roadmap. Current efforts focus on translating strategic planning into practical implementation steps, exploring pilot installations with early adopter facilities that could provide real-world operational experience and market validation.
Future development will be guided by market reception, partnership opportunities within the recycling sector, and practical considerations around manufacturing, support infrastructure, and deployment logistics. The technology foundation is established; next steps involve determining how and when to bridge from validated prototype to commercial deployment based on realistic assessment of market conditions and available resources.
Making plastic sorting automation accessible
Democratising technology requires more than developing new capabilities. It demands reimagining how users interact with complexity, designing systems that hide technical sophistication behind intuitive interfaces accessible to non-experts. CLARA demonstrates that automated plastic sorting can move beyond the exclusive domain of large industrial facilities with dedicated technical teams, becoming a practical tool for the broader SME recycling sector. This accessibility shift has the potential to unlock significant European recycling capacity, advancing circular economy objectives by removing technological barriers that have historically prevented smaller facilities from contributing at scale.