At the NVIDIA GTC 2025 conference, held in San Jose, Francesco Sarno, R&D Engineer at the Swiss Battery Technology Center (SBTC), was awarded the Best Poster Presentation Award.

SBTC's work, titled "Pioneering Automated EV Battery Recycling: Leveraging Synthetic Data and Reinforcement Learning for Sustainable Disassembly", co-authored with Jesse Isoaho, Özhan Özen, Raphael Rätz, Gabriel Hidalgo and Christian Ochsenbein, presents an innovative solution to overcome the challenges linked to end-of-life (EOL) electric vehicle (EV) batteries.

The research outlined the main obstacles to AI-based automated recycling of EOL EV batteries, in particular the scarcity of public datasets specific to EV batteries.

To overcome these challenges, the team created a synthetic dataset to train AI models in simulation, leveraging NVIDIA Omniverse's ray-tracing rendering capabilities and realistic physics-based modeling.

Within this virtual environment, computer vision models were trained for object detection, segmentation, depth estimation, 3D reconstruction and 6D pose estimation, alongside reinforcement learning agents tasked with robotic disassembly operations.

Using domain randomization techniques, the knowledge acquired in simulation was successfully transferred to real-world applications, integrated with ROS2, and demonstrated promising results.