ROBOTIC SORTING

The sorting station inside the confined, restricted shipping container poses a significant challenge for collision-free pose tracking of our robot arms. Several of our theses investigated solutions on how to overcome this problem.

The hydropower plant's location at the Limmat River provides realistic testing conditions and a source of actual river waste. During the Zurich Street Parade 2023, we collected and analyzed the trash that accumulated in the river during the one-day event.

The robotic sorting container we are currently developing, with built-in solar panels and batteries, will enable autonomous trash sorting even at remote locations. Built on a modular concept, we implemented our analysis and sorting technology into a single device. The system has been in action in Zürich since 2023!

We investigate reinforcement learning based strategies for pick-and-place maneuvers with multiple industrial manipulators. The goal is to increase sorting efficiency by maximizing material mass throughput for conveyor belt tasks.

NVIDIA's GPU-accelerated physics simulation enables massive parallelization. We aim to learn end-effector control and dynamic grasping strategies from scratch within minutes.