The Robotic Systems Lab designs machines, creates actuation principles, and builds up control technologies for autonomous operation in challenging environments.
We released our physics engine and a few RL examples. Check out our github repos below!
raisimLib (): simulator
raisimOgre (): visualizer
raisimGym (): gym examples
Excited to unveil Orbit v0.3 with many new features & improvements. Orbit is a modular robot learning framework that leverages the latest simulation capabilities and supports different learning paradigms! Join us in revolutionizing robotics! More info here
Check out
#ANYmal
running across gaps, slopes, and stepping stones as part of our submission "Perceptive Locomotion through Nonlinear Model Predictive Control".
This work is currently under review.
- video:
- preprint:
#robotics
Crafting multi-contact loco-manipulation behaviors for sparsely-defined tasks, like traversing spring-loaded doors or closing heavy dishwashers, is challenging. Check out our proposed solution for our legged mobile manipulator:
@SciRobotics
@ETH_en
We open source our fast GPU based elevation mapping software!
This includes features like traversability estimation and inpainting which are useful for navigation and legged locomotion.
It was used during the DARPA SubT Challenge by team
@CerberusSubt
.
Check out our
#ICRA2024
paper "ViPlanner: Visual Semantic Imperative Learning for Local Navigation"
introducing a novel planner that combines depth and semantic information.
Website:
Paper:
Full Video:
1. Legged systems have traditionally been controlled using trajectory optimization (TO). Such hierarchical model-based methods offer intuitive cost function tuning, accurate planning, and generalization. However, violations of modeling assumptions are common sources of failure.
Check our new IROS 2022 paper on "Locomotion Policy Guided Traversability Learning using Volumetric Representations of Complex Environments" by Jonas Frey, David Hoeller, Shehryar Khattak, Marco Hutter.
Video:
Paper:
@ETH_en
Congratulations to Vassilios Tsounis for the successful PhD defense on the application of RL for motion planning and control of quadrupeds!
@eth_dmavt
@rsiegwart
Check out our
#RSS2024
paper "Rethinking Robustness Assessment: Adversarial Attacks on Learning-based Quadrupedal Locomotion Controllers" to identify weak spots in the SOTA neural-net controllers.
Full video:
#legged_robot
#reinforcementlearning
RL controllers for quadruped locomotion are great but black-box. Are they safe enough? Our work in
#RSS2024
reveals the risk of small noises and low-frequency velocity commands falling SOTA controller, which helped win DARPA SubT Challenge, even on the simplest flat terrain.
Our paper "Whole-Body MPC and Online Gait Sequence Generation for Wheeled-Legged Robots" is accepted for
#IROS2021
in
#Prague
!
Check out the full video:
#Robotics
#Robots
Our new paper explores the use of parallel elastic actuators in quadrupedal robot design and control, with a focus on utilizing learning-based techniques. Results show promising improvements in efficiency.
#QuadrupedalRobotics
#Robotics
#DeepLearning
3. In our most recent work, we have combined TO and RL to synthesize a unifying control policy that leads to both accurate and robust locomotion (
#ScienceRoboticsResearch
,
@ETH_en
,
@ScienceMagazine
).
Full video:
Paper:
Start of the
@ETH_en
@eth_dmavt
robotics summer school 2021. 25 ETH and 25 international students learn in one week how to make an autonomous search and rescue robot.
#ARCHE
@swissrobotics
Congratulations to Giorgio for receiving the IEEE RAL best paper award! This work was done as part of the EU H2020 Project
@THING_H2020
. Check out the paper ()
RSL's field test engineers deployed two ANYmal C robots in the Seemühle underground mine as part of Team
@CerberusSubt
for the DARPA
#subtchallenge
A big thank you to Vale Schoch and Peter Schoch for making this possible!
Congratulations Dr. Fabian Jenelten for a successful defense. You managed to bridge
#MPC
and
#RL
to make
#ANYmal
a robust and versatile climber! Stay tuned for great videos to come soon.
@eth_dmavt
@ETH_en