Sarah Willis: "Automating waste sorting brings the benefit of transparency and traceability of the materials sorted"

During the FANUC Open House event, Rob Coker spoke with Sarah Willis, Head of Marketing for machine learning and recycling robotics specialist Recycleye about how AI and robotics & automation is revolutionising sorting technology.

In what ways does the Recycleye Vision AI technology help improve the sorting process?

Our low-cost, AI-powered Recycleye Vision system replicates the power of human vision, using advanced machine-learning algorithms to provide automatic, image-based detection of individual items in co-mingled waste streams.

This allows for 100% sampling of waste streams on a quality control line, transforming collected data into a dashboard in which plant managers can monitor their composition and granularity - bringing a previously impossible visibility to waste sorting. The vision system can also differentiate between classifications such as food-grade and non-food-grade plastics and packaging vs non packaging, prohibiting valuable recyclates from being downcycled.

Working with the Recycleye Robotics solution, this enables the automated picking of waste in materials recovery facilities that is faster, traceable and more accurate than human pickers.

The AI software looks pretty advanced. Can you tell us a little bit about how advanced it is?

Waste is perhaps the most complex computer vision problem – items are overlapping, crushed in many different ways and covered in dirt.

Recycleye had to develop a range of new algorithms in order to achieve the incredible +95% accuracies the system now provides (some of our work academic research has been published on our WasteNet platform). We are also the first to have brought the speed of the algorithm up to 60 frames per second meaning the system will, on average, classify each item 120 times on a conveyor belt (if it makes one error it will still have the average of the other 119 times) - this also means we are the only company with large AI models that can be installed on top of optical sorters.

At what speeds is it capable of accurately identifying and sorting waste streams?

Recycleye Vision and Recycleye Robotics picks at around 50 picks a minute with our 6 axis robot and around 70 picks with the delta robot on display at the Fanuc open days.

And how does it help out those manually sorting the waste?

The Recycleye Robotics solution can be used alongside human pickers, with a safety cage that ensures a safe distance is maintained between the hardware and people manually sorting waste.

Is the sorting/recycling industry slowly becoming reliant on AI and automation by effectively allowing it to 'do the dirty work’ for us?

Automating waste sorting brings the benefit of transparency and traceability of the materials sorted.  This will become increasingly important with the requirement for increased sampling with EPR.  Automated waste sorting is also more accurate, faster and can run 24/7 which is more economically efficient and safe than reliance on manual pickers, a role which experiences high turnover and recruitment issues.

In what ways has FANUC expertise contributed to realising this technology?

Our robotics solution was developed in collaboration with the team at Fanuc in the UK.  We now work with Fanuc on a European exclusivity basis for the manufacture and maintenance of our robotics.

Does collaboration with such partners not hinder Industry 4.0 uptake by elongating the value chain and time to market?

When robotics, AI and automation come as part of the machinery as in this case, does it help make the case for greater uptake of Industry 4.0 technology and software?

Absolutely, we are, for example, embedding IoT in everything we do. This means that clients can gather data on the performance of their facility on a live basis (based on the compositional data provided by the vision system), receive augmentations through software updates that will include the accuracy of the vision system or optimize the path planning of the robot. Future applications will also include predictive maintenance.

Reluctance regarding robotics and automation uptake is one of the major themes of the FANUC Open House event. What are your own thoughts on improving uptake across manufacturing and industry generally?

The waste management industry is defined by the traditional analog machinery which has served MRFs for several decades, therefore it is expected that many players are hesitant towards automation. However, we believe that increasingly stringent UK, European and global policies will require classification and sampling sizes to a degree only capable by AI and robotics.

Additionally, our solutions are retrofittable to existing machinery, allowing facilities to invest in the technology as a compliment, rather than substitute, and without need for expensive renovations. Hence, automation is becoming an increasingly attractive, and in some cases necessary, prospect for existing MRFs, rather than simply newly designed facilities.

In what ways is Recycleye working with partners further along the value chain (e.g. recyclers, local governments, environmental groups) to roll this technology out?

Recycleye is collaborating with waste management companies and local councils at MRFs on projects to increase vision capabilities, such as food grade vs non-food grade plastic differentiation, and packaging vs non packaging. Recycleye also believes that technology is key in driving alignment to ever changing industry policies, and therefore endeavour to work closely with policy-making organisations to enable an automated future for recycling.

Where else might we be able to see more Recycleye solutions live in action?

We have Robotic and Vision installations in the UK, France, Italy and Northern Ireland at the moment.