KIT Develops Adaptive Robotic System to Salvage Broken Industrial Robots

Researchers at the Karlsruhe Institute of Technology (KIT) in Germany have developed a robotic system capable of predicting defects in broken robots and disassembling them to salvage valuable components. The technology represents a shift toward automated recovery for industrial machinery, which has traditionally been difficult to recycle due to the unpredictable nature of physical damage.

Presented at the IEEE International Conference on Robotics and Automation (ICRA) 2026 in Vienna, the system utilizes CAD models and probabilistic planning—specifically Partially Observable Markov Decision Processes (POMDP)—to estimate how a robot is broken. By determining the degrees of freedom for each part, the system can adjust its disassembly strategy in real-time. The disassembler can switch methods based on observation; for instance, if a screw is found to be stuck, the robot can pivot from unscrewing to a destructive milling process to remove the part.

The Challenge of Robotic Waste

This development comes as the global footprint of industrial automation expands rapidly. There are currently over 4 million industrial robots in use worldwide, a number projected to grow to more than 16 million by 2030. As deployment scales to meet manufacturing demands, the industry faces a mounting problem of electronic waste and decommissioned machinery.

Traditional disassembly is often manual and inefficient because damage varies from one unit to another, making standardized automation nearly impossible. Jan Baumgärtner, a researcher at KIT's wbk Institute of Production Science, noted that while building a robot with new parts is straightforward, taking apart a broken one is unpredictable, stating, "We can imagine 100 ways that something can go wrong."

Toward a Circular Economy

By combining mathematical modeling of damage with adaptive manipulation, KIT's approach is a critical step toward a "circular economy" for robotics and electronics. Automating the extraction and replacement of broken parts reduces the reliance on new raw materials and lowers the overall environmental impact of the robotics industry. The ultimate goal is to make the repair process so inexpensive that it becomes more cost-effective than producing entirely new electronic devices.

Future Outlook

As the system moves from research to potential industrial application, the focus remains on refining the predictive algorithms to handle a wider variety of hardware failures. While the core technical framework has been validated, the scalability of these adaptive strategies across different robot brands and models remains a key area for further observation.