How AI could resolve longstanding detection issues for recyclers

In March, AI development SME Recycleye hosted a webinar featuring Technical Sales Manager Paloma Aldeguer, who presented an in-depth overview on how AI and robotics can be used in the sorting and recycling industry.

Recycleye is a growing technology company using advanced machine learning, computer vision, and robotics to commodify waste. It was founded in 2019 by graduates of Imperial College London, has an exclusive agreement with robotics manufacturer FANUC, and backing from Microsoft, and has received funding from the European Union and the UK Government. The company is currently focused on municipal household waste, working with clients in the UK, France, Italy, Germany and Australia.

Aldeguer began her address by introducing automation in waste sorting and explaining automated quality control. Key to this is a vision system powered by deep learning algorithms which focus on image recognition based on material type, as well as by shape and lustre. The vision system then powers the robotic unit which performs the separation of waste materials, including residuals, paper, HDPE and aluminium. Equipping robots with features such as a six-axis mechanism and optimising the algorithms ensure this technology can support the sorting process, according to Aldeguer

Aldeguer said: “Manual pickers naturally get tired over an eight-hour shift. With our vision systems monitoring this data, we were able to see that purity slowly decreases over time.”

The use of robotics resolved this challenge through consistency and highly reliable performances.

“In other words,” Aldeguer added, “there are no toilet breaks or holidays … 24/7 operations also result in 110 per cent more throughput, which means waste treatment more than doubles. If we think big and scale this up, we could absolutely double recycling rates.”

Recycleye has been working with Total, Valorplast and Citeo, in an attempt to identify food-grade PP with 97 per cent purity, which represents a clear example of where AI comes into its own. Equipping players to align purities with ever-changing legislation is crucial for robotic technology.

Recycling gets results

As a case in point, Aldeguer referenced the sampling scheme in the UK, which originally required 0.05% of the waste to be sampled. That’s 60kg out of 120 (metric) tonnes of waste. The new EPR scheme, however, has been published and the sampling requirement is now 60kg out of 75 tonnes. Introducing technology like Recycleye’s vision systems could enable 100 per cent traceability on the stream, surpassing the necessary requirements by far.

Further discussions with WRAP resulted in the key benefits of ramping up the capacity of sampling frequency without the need to find more space to store bales or find more manpower capacity, as well as the ability to achieve high accuracy on proportions and quantities. This could provide more visibility and enable producers to pay a fair amount for the materials.

All of this needs to be supported with high detection accuracies, without which there is no business model.

Aldeguer sees AI as complementary to technologies such as NIR as it is also able to detect black items, and base that detection on much more than material type alone. For example, the ability to sort PET trays from PET bottles, which conventional machinery is not capable of, and the capacity to sort aluminium cans from aluminium aerosols.

These are beginning to gather interest from materials recovery facilities (MRFs), according to Aldeguer.

Elaborating further on plastics sorting, Aldeguer highlighted the flexibility that this technology has. In many MRFs, lines are split in half. What the robot does involves cross-matching and separating HDPE jazz streams, as well as removing residuals from both sides: “The outcomes were absolutely smashing,” Aldeguer said. “We surpassed the expectations by achieving 99 per cent purity in HDPE. These outcomes absolutely justified the need for artificial intelligence in waste.”

Aldeguer reconfirmed that AI should remain complementary to other technologies such as optical sorters as Recycleye aims to push boundaries in purity, efficiency and granularity.