Vision-guided robotic system for automated stitch removal in end-of-life footwear disassembly
This paper presents a vision-guided robotic system that integrates a six-degree-of-freedom collaborative robot, YOLO-based machine vision, and force-controlled milling to automatically identify and remove stitches from end-of-life footwear, achieving sub-millimeter positioning accuracy to facilitate efficient material recovery and component reuse.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine a pair of shoes as a delicious, multi-layered sandwich. The top bun is the fancy leather upper, the bottom bun is the rubber sole, and the "glue" holding them together is a mix of sticky adhesive and a long, winding row of stitches around the edge. For years, recycling these sandwiches meant smashing them into a pulp, destroying the good parts. But what if we could carefully slice off the crust (the stitches) without squishing the filling? That's exactly what a team of researchers at Marche Polytechnic University tried to do with a robot that acts like a very precise, very patient chef.
The Robot Chef and Its "Eyes"
The star of the show is a six-armed robot (a collaborative robot, or "cobot") that doesn't just blindly follow a script. Think of it like a robot trying to trace a line drawn on a wobbly piece of paper. If the robot just followed a pre-drawn map (a CAD model) of a shoe, it would miss the stitches because real, used shoes aren't perfect; they are squished, stretched, and worn out.
To solve this, the robot has a camera mounted right next to its cutting tool, acting like a pair of eyes that look down at the work. Before it starts cutting, the robot takes a "measurement cycle" where it stops at every step of its path, snaps a photo, and asks a smart computer brain (an AI model called YOLOv9) to say, "Hey, the stitch is actually here, not where your map says it is!" The AI spots the stitches like a hawk spotting a mouse in a field, even if the lighting is a bit weird or the shoe is a bit crumpled. Once the AI gives the new coordinates, the robot adjusts its path and gets ready to cut.
The Cutting Dance
Once the robot knows where the stitches are, it enters the "removal cycle." Here, it uses a spinning milling tool (a tiny, high-speed drill bit) to slice through the thread. But it can't just push hard; if it pushes too hard, it might slice through the leather upper, ruining the part we want to save. If it pushes too soft, it won't cut the thread.
The robot uses a "force control" system, which is like holding a toothbrush while brushing your teeth. You don't just slam it against the wall; you feel the resistance and adjust your pressure. The robot does the same, gently pressing down with a specific amount of force (tested at 5, 10, and 15 Newtons) to ensure it cuts the thread but leaves the leather alone.
What Actually Happened?
The researchers tested this system on real shoes, specifically a right shoe with a white rubber sole and black leather upper. They ran the robot through four full cycles to see how well it worked.
The results were promising but not perfect. On the straight sides of the shoe, the robot was incredibly accurate, missing the stitch line by less than half a millimeter on average. It was like a tightrope walker staying perfectly centered. However, the robot initially struggled at the tricky curves, specifically the heel. In the heel area, the robot initially cut too deep, nicking the leather upper. In the toe region, the tool had moved excessively toward the sole, but after the team adjusted the camera position so the cutting tool was centered in the image, the errors in the toe became negligible, with the tool maintaining perfect contact with the stitching line.
The team found that the problem wasn't the robot's brain, but its "eyes." The camera was positioned in a way that made it hard to see the stitches clearly in those curved corners. After they moved the camera so the cutting tool was right in the center of the image, the errors dropped significantly, and the toe region performed well.
The Verdict
The study proves that this "robot chef" approach is feasible. The robot successfully removed stitches without destroying the shoe, with average positioning errors staying below 1 mm. However, the paper explicitly notes that the heel region still had some trouble, with errors peaking above 1.5 mm in some tests, which is just outside the safe zone for a 3 mm thick cutting tool.
The authors are careful to say this isn't a "solved" industrial problem yet. They point out that the AI model was trained on a relatively small dataset (300 images), which might be why it sometimes got confused in the complex curves. They also mention that while the robot is "collaborative" (meaning it's designed to work near humans), using a spinning cutter means we still need safety measures, like a laser sensor to slow the robot down if a human gets too close, rather than a full cage.
In short, the paper suggests that combining a robot, a camera, and a smart AI can successfully untangle the knots of old shoes, paving the way for a future where we can reuse the fancy leather parts of our sneakers instead of throwing them away. But for now, the robot still needs a little help with the tricky curves of the heel.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.