New legal requirements for textile collection have created a strained situation for municipalities and other stakeholders who cannot keep up with the increase in textile waste. Could automated pre-sorting be part of the solution? Automation researchers at University West are set to investigate this in a new project.
“Among the large amounts of collected textiles that come in, there are many items that don’t belong there – for example bags, toys, shoes, and other objects. Today, this pre-sorting is done manually, which is time-consuming, heavy, and monotonous work,” says Fredrik Danielsson, Professor of Automation at University West.
“In this first pre-sorting phase, we believe a smart robot could do part of the job. We will develop a robotic system that can identify items that should be sorted out and remove them from the conveyor belt. This requires, among other things, that the robot has advanced ‘hands’ that can grasp and lift objects. Right now, we are looking into what technology is needed for that.”
AI model trained with images
Fredrik and his colleagues will teach the robot to recognize odd objects. A large number of images of each type of item will be produced and used to train an AI model. Based on this model, the robot can determine whether something should be removed and, using coordinates, identify its position on the conveyor belt.
“An enormous amount of image material is needed for the AI model to have sufficient training data. By comparison, you need around 6,000 images of a screw – from different angles, lighting conditions, and backgrounds – for a robot to be able to identify and pick it up.”
“To gather enough image material, we can partly make use of open-source resources, and we will also receive images from the test and demo facility Wargön Innovation, one of our collaboration partners. The aid organization Human Bridge is also part of the project.”
Robot to be tested on a conveyor belt
So far, the research team has taught the robot to open the large plastic bags in which the textiles arrive. They will also build a simulated conveyor belt in the robotics lab on campus to test the robot’s abilities to identify and pick up objects.
“Our goal is for the robot to be able to recognize and remove 80 percent of all non-textile materials.”
A complex task
The project will run for two years, a short time frame for developing a robot for this type of task.
“It is a very complex assignment for a robot. But with the rapid development ongoing in AI-driven robotics, new possibilities are constantly emerging.”
“Hopefully, we can reach a point where smart robots can assist in the initial phase of textile sorting and thereby contribute to efficient textile reuse later in the process.”
“If this succeeds, the next step could be to assess whether robots can help with sorting based on the condition and quality of the textiles. That task is considerably more difficult for robots with the technology we have today. For now, humans are unmatched.”
Contact: Fredrik Danielsson, Professor of Automation, University West
FACTS
• The research project EcoSort is funded by Formas and runs from December 1, 2025, to December 2027.
• A new EU law went into effect on January 1, 2025, requiring member states to introduce systems for separate collection of textile waste for reuse or recycling. Sweden is the first to implement the law.
• Watch a video of the robot opening plastic bags containing textiles.
At University West, we conduct research in collaboration with society to help create a better future.