EU’s ReconCycle develops self-reconfiguring technology with e-waste disassembly robot

The EU research project ReconCycle developed a robotic work cell to automate the disassembly of electronic waste (e-waste). The project published an overview of its results on December 20, 2024, and detailed results were available on CORDIS as of May 26, 2025. The work cell integrates modular hardware, soft robotics and AI, and can self-reconfigure to handle a variety of device models and damage conditions.

ReconCycle’s work is a step toward automating e-waste recycling and could help reduce environmental impact and improve worker safety. The integration of soft robotics and AI also has potential applications in other industries.

ReconCycle’s work marks progress in automation in the e-waste recycling sector. E-waste disassembly has traditionally relied on manual labor, and flexible automation that can handle a variety of models and damage conditions has been a challenge. ReconCycle addressed this challenge by combining modular hardware and AI. In particular, self-reconfiguration enables the system to adapt automatically when device types or models change. This is a major advance over conventional industrial robots, which operate using fixed programs. A key technical feature of the project is the integration of soft robotics and AI. Soft grippers such as the qb SoftHand 2 and VSG make it possible to handle fragile components without damaging them, while AI-based action prediction automates complex disassembly procedures. In particular, the Action Predictor module, which combines a vision-language model (VLM) with retrieval-augmented generation (RAG), enables context-dependent action prediction and improves disassembly efficiency. For the industry, ReconCycle’s technology could help reduce costs and improve safety in e-waste recycling. Automating processes that rely on manual labor can reduce labor costs and keep workers away from hazardous tasks. The modular design also makes it easier to adapt to different devices, increasing the flexibility of recycling facilities. One unresolved issue is the challenge of scaling up for industrial deployment. The project has reached TRL 6 (demonstration stage), but further validation is needed before commercialization. Further research is also needed on AI training data and the limits of the system’s adaptability to different devices.

ReconCycle is built around standardized modules and includes torque-controlled manipulators, soft grippers (qb SoftHand 2, the variable-stiffness gripper VSG and SoftClamp), a cutter module, CNC milling and a tool-changing mechanism. Plug-and-Produce (PnP) connectors provide mechanical stability, electrical and pneumatic supplies, and data connections, enabling rapid reconfiguration. The software is ROS-based and uses the FlexBE behavior engine, Rviz and the Gazebo simulation environment. Its AI technology integrates computer vision and a vision-language model (VLM) to predict actions based on context. The project conducted tests against specified KPIs and confirmed improvements in adaptability, efficiency and safety.