Embedded Files
Matterix: Building Digital Twins for Robotics-Assisted Chemistry Labs
Description: This tutorial introduces Matterix, a GPU-accelerated simulation framework for building high-fidelity digital twins of chemistry laboratories. Participants will learn how to model and simulate robotic manipulation in lab environments, including interactions with liquids, powders, and laboratory equipment, as well as key processes such as heat transfer and basic reaction dynamics. We will cover how to construct digital twin environments using open-source assets and how to design and evaluate workflows through hierarchical planning and modular skill libraries. The tutorial will demonstrate how digital twins can be used to prototype and test robotic workflows in silico, enabling faster iteration, and supporting sim-to-real transfer. By the end of the session, participants will be able to use digital twins to design, evaluate, and improve robotics-driven laboratory systems.
Duration: 30 Minutes
Led by: Kourosh Darvish (co-organizer of this workshop)
Benchmarking Robot Systems for Laboratory Assistance: A Fitness-Based Approach
Benchmarking Robot Systems for Laboratory Assistance: A Fitness-Based Approach
Description: Effective deployment of robotic systems in laboratory environments requires benchmarking methodologies that go beyond traditional morphology-based classifications. This seminar presents a process-driven benchmarking framework that evaluates robots based on their fitness to perform specific tasks, particularly those involving physical interaction and human collaboration. The approach follows a structured pipeline: laboratory tasks are analysed, decomposed into fundamental operations, and translated into process quality requirements, which are then mapped onto quantitative robot performance metrics. These include both classical motion metrics such as accuracy and repeatability and interaction-centric metrics, covering force sensing, force control, contact reaction, safety, and manual guidance. By integrating these dimensions, the framework enables objective, reproducible comparison across diverse robotic platforms and highlights substantial differences in their suitability for delicate, contact-rich tasks typical of laboratory assistance. This fitness-based benchmarking further supports the classification of robots into functional categories aligned with their capabilities.
Ultimately, this methodology provides a principled foundation for selecting, evaluating, and improving robotic systems in laboratory contexts, enabling more reliable, safe, and application-specific deployment of robotic assistants.
Duration: 30 Minutes
Led by: Robin Kirschner (Technical University of Munich, MIRMI)
Page updated
Google Sites
Report abuse