On the Greek island of Kefalonia, the arithmetic of recycling has never worked. Too few people spread across too much geography. Too expensive to collect, too costly to sort, and then there's the matter of a ferry ride to get mixed waste to the mainland—burning diesel the whole way—only to have it picked over with mediocre results.
By early summer last year, something unusual appeared next to the Pallosti landfill: a stack of shipping containers housing five robots, running double shifts and sorting seven types of material without anyone building so much as a foundation.
This is RECLAIM's pitch, more or less. Take automated waste sorting—the kind rapidly spreading through billion-dollar facilities in North America and Europe—and cram it into a box you can truck to places the economics have left behind. The system can be running in under 48 hours, project partners say, and reconfigured in half an hour if the waste stream changes. It's a concept born from a straightforward question posed by a ten-partner European Union consortium: what if the robots went to the waste, instead of the other way around?
The Horizon Europe-funded effort wrapped up in August 2025 with €2,934,800 in EU backing. What it left behind is a working prototype and a bet that decentralization might solve what centralization can't.
The Portability Play
The portable robotic Material Recovery Facility—prMRF in the inevitable acronym—lives inside standard shipping containers. The design brief was practical to the point of bluntness: serve populations around 20,000, the kind of island towns, mountain municipalities, seasonal ports and festival grounds where waste volumes swing wildly or geography makes permanent plants a non-starter.
Setup involves dropping the containers, plugging into power (with partial solar assist built in), and firing up web-based monitoring software. The Kefalonia deployment ran two days a week under remote oversight from consortium partners: FORTH-ICS, the Greek research institute coordinating the project; ION, the local waste authority; and HERRCO, Greece's recovery organization.
That containerized form is what distinguishes this from the direction the industry has been heading. Companies like AMP Robotics are building fully automated "AMP ONE" systems for Waste Management. Glacier has robot fleets working Recology facilities in Seattle. Those are permanent installations, multi-million-dollar capital commitments designed for scale and anchored in concrete.
RECLAIM went the other direction. Partners including Spain's AIMPLAS and IRIS Technology Solutions, Germany's AXIA Innovation, and Belgium's KU Leuven designed for mobility, not maximum throughput. Whether that trade-off pencils out commercially is the open question.
What's Inside the Box
Five robots—called "Robotic Recycling Workers," or RoReWos, because European research projects live for this sort of thing—do the picking. The design skewed toward low cost and task-specific builds rather than repurposing off-the-shelf industrial arms. Pneumatic actuators, interchangeable grippers. The consortium claims an average of 120 picks per minute across the fleet, sorting PET, HDPE, PP, PS, aluminum, paper and carton packaging, and ferrous metals. Under demonstration conditions, project documentation from August 2025 put that at over 6,500 items per hour.
Classification happens through dual imaging: standard RGB computer vision paired with hyperspectral imaging from IRIS Technology. The HSI system analyzes material composition on the belt in real time—a technique the project partners say achieved over 94 percent material classification accuracy during development. A separate claim of 98 percent waste classification accuracy surfaced in trade press, though the methodologies behind the differing figures weren't laid out in public materials. (The gap between those numbers is a reminder that accuracy metrics in waste sorting can be slippery, depending on what exactly you're measuring.)
KU Leuven contributed the scheduling and grasp-planning algorithms that keep five robots from getting in each other's way, reducing cycle times when they're all picking from the same belt.
What Kefalonia Showed
The island pilot was designed to measure operational impacts beyond pure sorting speed. Project-end materials from August claimed approximately 1.6 tonnes of CO2 saved daily through reduced transport—a 36 percent cut in fuel consumption by eliminating mainland sorting trips—and 12 percent faster processing via on-site separation.
An earlier metric from January, before the island deployment, estimated 1.2 tonnes of material recovered per shift with 0.7 tonnes of CO2 reduction per shift. Shift duration and waste composition weren't specified, and those kinds of details matter when you're trying to model real-world economics.
HERRCO handled on-site validation: checking purity, composition, efficiency during the demonstration phase. The University of Malta added a citizen-science angle—a "Recycling Data Game" mobile app that crowdsourced waste annotations to help train the AI models. It won a transnational collaboration award, which is the sort of thing that looks good in grant reports.
What Happens Next
The EU project formally closed August 31, 2025. Commercialization is now the question that matters. Project partners have floated interest from new airport developments—Heraklion was mentioned—along with mountain regions and large events where temporary or seasonal waste surges happen. The argument is straightforward: decentralized recovery works where centralized processing doesn't, at least not economically.
The prMRF isn't alone in trying to crack the waste-sorting problem with robots. But portability and plug-and-play deployment for underserved markets? That's a niche the big North American automation players haven't chased, perhaps for good reason. Fixed facilities scale. Mobile ones are a harder sell.
Whether the containerized model finds commercial traction will come down to hard math: can municipalities, event operators, and remote infrastructure developers justify the capital and operating costs against avoided transport expenses and faster local turnaround? RECLAIM's wager is that for certain geographies—the islands, the isolated, the seasonal—recovery in a box beats no recovery at all.
It's a reasonable bet. Whether it's a winning one is another matter entirely.
