Researchers at the Karlsruhe Institute of Technology (KIT) have developed a robotic disassembly system that can autonomously take apart broken robots and devices. The system uses a predictive algorithm combined with CAD models to guess how a device is broken, then directs robotic manipulators to disassemble it while protecting valuable components. It continuously checks results against predictions and adapts — for example, switching from unscrewing to milling when a screw is stuck. The goal is to enable a circular economy where broken devices are repaired and repurposed rather than discarded, ultimately making repair cheaper than manufacturing new products. The research was presented at ICRA 2026 in Vienna.
Questions this post answers
How does the KIT robotic disassembly system handle unexpected failures like a stuck screw during disassembly?
The KIT system continuously compares observed part behavior against its predictions and adapts its strategy in real time. If a screw remains in place after an unscrewing attempt, the system detects the deviation and switches to an alternative method — such as milling away material — to remove the part. This avoids wasted effort and protects components that need to be salvaged intact. Engineers working on adaptive robotics track disassembly research like this on daily.dev.