Robots Use Probabilistic Planning to Overcome Surprises When Dismantling Machines

Taking apart old or damaged machines is far messier than building new ones. Traditional industrial robots work best when every screw turns and every component is exactly where the blueprint says it should be, so unexpected problems can stop the whole process.

To tackle this issue, researchers at the Karlsruhe Institute of Technology (KIT) in Germany, led by Jan Baumgärtner and his team, have created a new robotic disassembly system.

The system relies on a type of probabilistic planning called a Partially Observable Markov Decision Process (POMDP). Instead of assuming it knows everything, the robot assigns probabilities to different possible problems and updates those beliefs as it gathers new information.

Because of this approach, the robot can detect obstacles such as stuck screws, missing components, or even machines that no longer match their original designs. When it encounters a problem, it can switch tools – for example, using a milling tool to cut around a stubborn screw – and continue the task.

In physical experiments, the probabilistic planner achieved faster disassembly times than traditional deterministic planning whenever uncertainty was present. The advantage was especially clear when the robot had alternative routes to reach the target part.

The research was presented at the 2026 IEEE International Conference on Robotics and Automation in Vienna, highlighting the practical potential of the method.

If robots can reliably handle the unpredictability of real‑world repairs, manufacturers could recycle more components and make the whole disassembly process cheaper and greener.

COPYRIGHT DISCLAIMERS

4English does not own the rights to this content. All rights belong to the owner. 4English only uses this content to help users learn English via the app's features. Contact to request the removal of your content: contact@4englishapp.com.

Praticar Escuta