Engineers have designed a new software controller trained to feel out and react to the environment that successfully guided blind 4-legged robots through snow, rubble, slippery streams, and other challenging terrain. Learn more in Science Robotics: ($) fcld.ly/zfhr3ro

Learning a navigation policy in dynamic environments in 10 minutes and deploying it sim-to-real is not as hard as it sounds. See our RA-L paper arxiv.org/abs/2103.04351 with David Hoeller, Lorenz Wellhausen, Farbod Farshidian, Marco Hutter, youtu.be/CICWcLJ3aPs ⬇️ Key ingredients ⬇️

iMAP is a new way to do SLAM: we learn an implicit neural representation *in real time* and track an RGB-D camera against it. The implicit map fills holes; completes the unseen backs of objects; and maps a whole room in only 1MB of weights. From the Dyson Robotics Lab, Imperial.

Personally I feel very sad about this article. First I want to say upfront that I have no issue with the original paper from the amazing researchers from Berkeley. My complains only apply to this article.

Our #NVIDIA Isaac Gym detailed technical report and benchmark results are now up on arXiv. Supercharge the training of your Reinforcement Learning (RL) networks by harnessing the power of the #GPU. Check out more details here: nvda.ws/3mD2HcO

Our team CERBERUS SubT won the DARPA #SubTChallenge!!! #Robotics is becoming real!

Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning A video from their paper of thousands of robots learning to walk within minutes in a simulation on a single GPU. Near the end, they transfer the policy to the actual robot. arxiv.org/abs/2109.11978

I’m super excited to see ideas from complex systems such as swarm intelligence, self-organization, and emergent behavior gain traction again in AI research. We wrote a survey of recent developments that combine ideas from deep learning and complex systems: arxiv.org/abs/2111.14377

Our stuff was shown in NVIDIA GTC

If you missed the humanoid-quadruped transformer at NVIDIA's #GTC, then you can watch the full video here: youtu.be/kEdr0ARq48A #Robotics #reinforcementlearning #nvidia #transformer #swissmile ETH Zurich NVIDIA Omniverse NVIDIA GTC

Just pushed version 0.2.0 of Warp. Adds support for multidimensional arrays, slices, additional NVDB formats, and better support for differentiating through dynamic loops: github.com/NVIDIA/warp

Whoa. Learned behavior from end-to-end training that combines navigation and locomotion (w/ motion capture for state estimation) by Robotic Systems Lab. Paper at IROS2022 next month, more here: sites.google.com/leggedrobotics…

Interested in pushing the limits of legged robots? Check out how ANYmal learns to jump, climb and run in our @iros202 paper: "Advanced Skills by Learning Locomotion and Local Navigation End-to-End". Nikita Rudin, David Hoeller, Marko Bjelonic youtube.com/watch?v=Xoe8a_…

Mapping the terrain for locomotion is hard for quadrupedal robots. The camera setup results in blind spots and the state estimator suffers from drift.We propose a solution in our RA-L IROS2022 paper youtu.be/3zsvqCrztLg David Hoeller Nikita Rudin @ChrisChoy208 Prof. Anima Anandkumar

Roboticists from Robotic Systems Lab and NVIDIA Embedded are teaching four-legged robots climb and jump. After training in simulation, the robots can autonomously decide how to scramble over and under obstacles, which will help them do dangerous jobs so that humans don't have to.

🔥Exciting news 🤖 Our latest research by David Hoeller, Nikita Rudin, Eris in Science Robotics unlocks new achievements:  Unprecedented agility in quadrupedal robots, mastering locomotion, navigation, and perception through deep reinforcement learning! NVIDIA Robotics

I may have missed recent progress, but that's extremely impressive. Definitely the best quadruped robot motion I have ever seen, by far. Robotic Systems Lab youtube.com/watch?v=PjWvf9…

🔥 Exciting news 🤖 Our latest research by David Hoeller, Nikita Rudin, Eris in Science Robotics unlocks new achievements: Unprecedented agility in quadrupedal robots, mastering locomotion, navigation, and perception through deep reinforcement learning! NVIDIA Robotics