Robotic Systems Lab

Precision throwing with a 12-ton hydraulic machine. Our latest research leverages Reinforcement Learning to teach heavy construction machinery how to dynamically extend its reach. We'll preset the details at #IROS2024 in Abu Dhabi, stay tuned! Meanwhile, you can check out the video and our paper: 🎥 https://lnkd.in/dDm-MuNy 📄 https://lnkd.in/drb2ud9f

The Robotic Systems Lab participated in the Advanced Industrial Robotic Applications (AIRA) Challenge at the ACHEMA 2024 process industry trade show, where teams demonstrated their teleoperated robotic solutions for industrial inspection tasks. We competed with the ALMA legged manipulator robot, teleoperated using a second robot arm in a leader-follower configuration, placing us in third place for the competition! Full video: https://lnkd.in/gavMCasZ

RL controllers for quadruped locomotion are great but black-box. Are they safe enough? Our new work in #RSS2024 reveals the risk of small noises and low-frequency velocity commands falling SOTA controller, which helps win SubT Challenge, even on the simplest flat terrain. We explored whether standard robustness tests, human experiences, and domain randomization are enough, together with tons of other extensive studies. This work is collaborated with Chong Zhang, Takahiro Miki, Joonho Lee, Marco Hutter, Stelian Coros at Robotic Systems Lab, Computational Robotics Lab - ETH Zürich, and ETH AI Center. Project page: https://lnkd.in/eHKhfvPZ

Floating Robotics is very cool spin-off of Robotic Systems Lab automating precise manipulation in farming. Vote for us to go first in #TOP100SSU engineering vertical 🚀

https://lnkd.in/dcMxDzZB

#ICRA24 is over with an incredible participation from RSL. After many insightful conversations, sharing of ideas and discussion of the current research we are going back to work to address the future challenges 🚀 Feel free to reach out to us, check the presented work and possible future works at https://rsl.ethz.ch/

Meet us from May 13 to 17 at #ICRA2024! Here are all our contributions: Mon May 13 Workshops ==================== a) Can robots rival animal agility by learning on their own? (Based on the recent ANYmal Parkour work https://lnkd.in/e6mtBCdW). Agile Movements II workshop. 11:25 - 11:50. Room F205 b) Cybathlon - Legged Mobile Assistance for Quadriplegics, Poster at 15:50 - 16:20, Assistive Systems Workshop. Room G402. Tue May 14 ========== a) Resilient Legged Local Navigation: Learning to Traverse with Compromised Perception End-To-End, 10:30 - 12:00, Room CC - 301, https://lnkd.in/eNFYrUAw b) SpaceHopper: A Small-Scale Legged Robot for Exploring Low-Gravity Celestial Bodies, 13.30 - 15.00, Room AX-205, https://lnkd.in/g85qnBeK c) ViPlanner: Visual Semantic Imperative Learning for Local Navigation, 16:30 - 18:00, Room CC-419, https://lnkd.in/gc6SN6HW d) Pedipulate: Enabling Manipulation Skills using a Quadruped Robot’s Leg, 16.30 - 18.00, Room CC-311, https://lnkd.in/gXZpR3hd e) X-ICP: Localizability-Aware LiDAR Registration for Robust Localization in Extreme Environments, 16:30 - 18:00, Room NT - G402, https://lnkd.in/gFg8aurv Wed May 15 ========== a) Symmetry Considerations for Learning Task Symmetric Robot Policies, 10:30 - 12:00, Room CC - 416 b) Learning to Walk in Confined Spaces Using 3D Representation, 13:30 - 15:00, https://lnkd.in/gSrMMsZp c) Seeing Through the Grass: Semantic Pointcloud Filter for Support Surface Learning, 13:30 - 15:00, Room AX - F205, https://lnkd.in/gheRAF4k d) SMUG Planner: A Safe Multi-Goal Planner for Mobile Robots in Challenging Environments, 16:30-18:00, Room CC-303, https://lnkd.in/gqT9PQAY e) Covariance Based Terrain Mapping for Autonomous Mobile Robots, 16:30 - 18:00, https://lnkd.in/ggxXx4ey f) Learning Risk-Aware Quadrupedal Locomotion using Distributional Reinforcement Learning, 16:30 - 18:00, Room AX - F205, https://lnkd.in/dZtXwYVy

LunarLeaper is a robotic mission aiming to expire lava caves on the Moon 🌝 🌙 We discussed some aspects of our mission with the Swissinfo team. Thanks for coming by! #totheMoon with Valentin Bickel, Philip Arm, Simon Stähler and Hendrik Kolvenbach 🇪🇺🚀#ESA 🇨🇭 #eth Department of Earth Sciences, ETH Zurich

Harnessing the power of data-driven algorithms has revolutionized deep learning, but acquiring real-world robotics data remains expensive and time-consuming. What if we could overcome this hurdle by tapping into cutting-edge simulation technologies to bridge the simulation and reality gap? We are excited to share Orbit, a unified framework for robot learning that capitalizes on the capabilities of NVIDIA Isaac Sim. Orbit not only integrates insights from our research on legged robots but also offers a flexible and robust tasking framework, making it easy to integrate with your robot! It directly supports various learning paradigms – reinforcement learning, imitation learning, motion planning – so you can focus on solving hard and complex robotic tasks. Additionally, as announced at NVIDIA GTC, Orbit's development has joined forces with the NVIDIA Isaac ecosystem, evolving into Isaac Lab. This transition brings about a leaner simulation application tailored specifically for robot learning. More exciting news coming soon! Want to kickstart your robot learning project? Check out our paper and repository to see what Orbit can offer for your project. Paper: https://lnkd.in/g27deX37  Documentation: https://lnkd.in/ewRfqrpz  Code: https://lnkd.in/eyNDHDg2 A big shoutout to our incredible team (Mayank Mittal, Nikita Rudin, David Hoeller, Pascal Roth), partners from The AI Institute and NVIDIA Robotics, and the open-sourced community for their active involvement in the framework’s development! #Robotics #AI #Simulation #NVIDIA

We are happy to release the open-source extension for our Wild Visual Navigation framework (Paper + Code). Title: "Wild Visual Navigation: Fast Traversability Learning via Pre-Trained Models and Online Self-Supervision" Authors: Matías Mattamala Jonas Frey Piotr Libera Nived Chebrolu Georg Martius Cesar Cadena Marco Hutter Maurice Fallon Example Video: https://lnkd.in/eWjBH_ih New Paper (under review): https://lnkd.in/e6c9Ga8t GitHub: https://lnkd.in/e3rsT9du Project Page: https://lnkd.in/eM_9eK5t Mode Videos: https://lnkd.in/eM_9eK5t RSS Paper: https://lnkd.in/ewxQ_4i2 A collaboration of the University of Oxford, Oxford Robotics Institute, Max Planck Society, Max Planck Institute for Intelligent Systems, and ETH Zürich. 165 Issues - 724 Commits - Uncountable Coffees