Researchers at the Karlsruhe Institute of Technology in Germany have developed a robotic system that can autonomously dismantle broken products while protecting valuable components. The technology could support a circular economy in which damaged machines and electronics are repaired or rebuilt rather than discarded, tells IEEE Spectrum.
Automated disassembly is harder than automated manufacturing because broken products are unpredictable. A manufacturing robot follows predefined steps using undamaged components. A recycling robot, however, must determine what has failed and adapt its actions accordingly. KIT’s system combines a predictive algorithm with robotic manipulators capable of unscrewing, handling, and even milling components. It continually compares actual results with its predictions and changes its strategy when necessary.
The process begins with CAD models of the product and its individual parts. These models tell the system how components should normally move. The robot then nudges parts and analyzes their degrees of freedom. Unexpected movement can indicate problems such as loose screws, while restricted movement may suggest corrosion or deformation. A mathematical model uses these observations to predict damage and refine the disassembly plan.
The system can also prioritize components that users want to recover intact. If it encounters an unexpected obstacle, such as a stuck screw, it can abandon the original approach and choose another removal technique. This adaptability helps avoid unnecessary work and damage to reusable parts.
Researcher Jan Baumgärtner ultimately envisions factories containing many specialized robotic arms that dismantle products instead of assembling them. Such systems could automatically identify failures, remove damaged components, install replacements, and rebuild devices.
The goal is economic as well as environmental. If automated disassembly makes repairing existing products cheaper than manufacturing replacements, robotics could help reduce waste while recovering valuable components. The research was presented at the 2026 IEEE International Conference on Robotics and Automation in Vienna.
