- Видео 133
- Просмотров 883 928
Dynamic Robotics Laboratory
Добавлен 7 апр 2011
The members of the Dynamic Robotics Laboratory seek to achieve agile locomotion for legged robotic systems. Through creating robots that move swiftly and can easily handle impacts and kinetic energy transfer, we hope to push the limits of bipedal robots to make them useful in the real world. Examples include running, skipping, hopping, and walking up/down stairs. Many of these tasks are difficult due to hybrid, nonlinear dynamics caused by contacts, and ambiguous reward specification. To achieve these goals, our research aims to combine the first principles of legged locomotion with learned control systems.
Our focus is on both the control hierarchy as a whole as well as the individual low level components. For each piece, we want to identify the correct structure based on our knowledge of dynamics and apply learning to solve individual difficult problems.
Our focus is on both the control hierarchy as a whole as well as the individual low level components. For each piece, we want to identify the correct structure based on our knowledge of dynamics and apply learning to solve individual difficult problems.
Learning Decentralized Multi-Biped Control for Payload Transport
Project page: decmbc.github.io/
Paper: arxiv.org/abs/2406.17279
Abstract - Payload transport over flat terrain via multi-wheel robot carriers is well-understood, highly effective, and configurable. In this paper, our goal is to provide similar effectiveness and configurability for transport over rough terrain that is more suitable for legs rather than wheels. For this purpose, we consider multi-biped robot carriers, where wheels are replaced by multiple bipedal robots attached to the carrier. Our main contribution is to design a decentralized controller for such systems that can be effectively applied to varying numbers and configurations of rigidly attached bipedal robots without retrainin...
Paper: arxiv.org/abs/2406.17279
Abstract - Payload transport over flat terrain via multi-wheel robot carriers is well-understood, highly effective, and configurable. In this paper, our goal is to provide similar effectiveness and configurability for transport over rough terrain that is more suitable for legs rather than wheels. For this purpose, we consider multi-biped robot carriers, where wheels are replaced by multiple bipedal robots attached to the carrier. Our main contribution is to design a decentralized controller for such systems that can be effectively applied to varying numbers and configurations of rigidly attached bipedal robots without retrainin...
Просмотров: 1 256
Видео
Revisiting Reward Design and Evaluation for Robust Humanoid Walking
Просмотров 1,8 тыс.4 месяца назад
Project page: b-vm.github.io/Robust-SaW/ Paper: arxiv.org/abs/2404.19173 A necessary capability for humanoid robots is the ability to stand and walk while rejecting natural disturbances. Recent progress has been made using sim-to-real reinforcement learning (RL) to train such locomotion controllers, with approaches differing mainly in their reward functions. However, prior works lack a clear me...
Sim-To-Real Learning for Humanoid Box Loco-Manipulation
Просмотров 7895 месяцев назад
Supplementary video for 2024 IEEE International Conference on Robotics and Automation arXiv: arxiv.org/abs/2310.03191 Abstract - In this work we propose a learning-based approach to box loco-manipulation for a humanoid robot. This is a particularly challenging problem due to the need for whole-body coordination in order to lift boxes of varying weight, position, and orientation while maintainin...
Learning Vision-Based Bipedal Locomotion for Challenging Terrain
Просмотров 1,8 тыс.5 месяцев назад
Supplementary video for 2024 IEEE International Conference on Robotics and Automation arXiv: arxiv.org/abs/2309.14594 Abstract - Reinforcement learning (RL) for bipedal locomotion has recently demonstrated robust gaits over moderate terrains using only proprioceptive sensing. However, such blind controllers will fail in environments where robots must anticipate and adapt to local terrain, which...
Learning Dynamic Bipedal Walking Across Stepping Stones
Просмотров 1,3 тыс.Год назад
This is the submission video for the 2022 IROS (IEEE International Conference on Intelligent Robots and Systems) paper "Learning Dynamic Bipedal Walking Across Stepping Stones" by Helei Duan, Ashish Malik, Mohitvishnu S. Gadde, Jeremy Dao, Alan Fern, Jonathan Hurst Preprint link to full paper: arxiv.org/abs/2205.01807 Abstract: In this work, we propose a learning approach for 3D dynamic bipedal...
Kevin Green's Robotics PhD Thesis Defense
Просмотров 2,3 тыс.Год назад
"Agile Bipedal Locomotion via Hierarchical Control by Incorporating Physical Principles, Learning, and Optimization" presented on August 31st 2022. Full written thesis available: ir.library.oregonstate.edu/concern/graduate_thesis_or_dissertations/g732dh894
Sim-to-Real Learning of Footstep-Constrained Bipedal Locomotion
Просмотров 1,3 тыс.2 года назад
This is the submission video for the 2022 ICRA (IEEE International Conference on Robotics and Automation) paper "Sim-to-Real Learning of Footstep-Constrained Bipedal Locomotion" by Helei Duan, Ashish Malik, Jeremy Dao, Aseem Saxena, Kevin Green, Jonah Siekmann, Alan Fern and Jonathan Hurst Preprint link to full paper: arxiv.org/abs/2203.07589 Abstract: Recently, work on reinforcement learning (...
Sim-to-Real Learning for Bipedal Locomotion Under Unsensed Dynamic Loads
Просмотров 4,4 тыс.2 года назад
This is the submission video for the 2022 ICRA (IEEE International Conference on Robotics and Automation) paper "Sim-to-Real Learning for Bipedal Locomotion Under Unsensed Dynamic Loads" by Jeremy Dao, Kevin Green, Helei Duan, Alan Fern, Jonathan Hurst Preprint link to full paper: arxiv.org/abs/2204.04340 Abstract: Recent work on sim-to-real learning for bipedal locomotion has demonstrated new ...
Cassie Robot Runs Turf 5k (Uncut)
Просмотров 55 тыс.3 года назад
To encourage us to push the limits of reliability, energy efficiency and speed we attempted to run a 5k (5 kilometers, 3.1 miles) with our bipedal robot Cassie. Our final goal was to run a 5k through the Oregon State University campus. That can be seen here ruclips.net/video/dY57qnD_O7U/видео.html This turf 5k was a test before we attempted the campus run. We finished with a time of 43:59.53 wh...
ICRA 2021: Learning Task Space Actions for Bipedal Locomotion
Просмотров 1,3 тыс.3 года назад
This is the presentation for our ICRA 2021 paper "Learning Task Space Actions for Bipedal Locomotion" by Helei Duan, Jeremy Dao, Kevin Green, Taylor Apgar, Alan Fern, Jonathan Hurst. Preprint: arxiv.org/abs/2011.04741 Supplemental Video: ruclips.net/video/8OCOzPqZcGM/видео.html
ICRA 2021: Sim-to-Real Learning of All Common Bipedal Gaits via Periodic Reward Composition
Просмотров 2 тыс.3 года назад
This is the presentation of our ICRA 2021 paper. The authors are Jonah Siekmann and Yesh Godse (who contributed equally), Alan Fern, and Jonathan Hurst. ArXiv Preprint: arxiv.org/abs/2011.01387 Paper's Supplemental Video: ruclips.net/video/4DnxV9lko_U/видео.html Additional Video of Hardware Results: ruclips.net/video/Wb0tIWBrjmc/видео.html
ICRA 2021: Learning Spring Mass Locomotion
Просмотров 2,2 тыс.3 года назад
This is the presentation of our 2021 ICRA and RA-L paper "Learning Spring Mass Locomotion: Guiding Policies With a Reduced-Order Model" by Kevin Green, Yesh Godse, Jeremy Dao, Ross L. Hatton, Alan Fern, Jonathan Hurst. IEEE Paper Link: ieeexplore.ieee.org/abstract/document/9380929 ArXiv Link: arxiv.org/abs/2010.11234 Supplemental Video: ruclips.net/video/80oJeaAd8CE/видео.html
Reliability Tests of a Learned Stair Climbing Controller
Просмотров 13 тыс.3 года назад
To show how reliable our learned controller is, we wanted to publish an uninterrupted video of ten stair ascents and ten stair descents. The controller being tested is described in our 2021 Robotics: Science and Systems paper "Blind Bipedal Stair Traversal via Sim-to-Real Reinforcement Learning" Preprint: arxiv.org/abs/2105.08328
Robot Learns to Climb Stairs Blind
Просмотров 39 тыс.3 года назад
We successfully used reinforcement learning to train a recurrent neural network to control our bipedal robot Cassie to climb stairs without any perception sensors such as LIDAR or cameras. This video accompanies our submission to the 2021 Robotics: Science and System conference. A preprint of the full paper can be found here arxiv.org/abs/2105.08328. "Blind Bipedal Stair Traversal via Sim-to-Re...
Learning Task Space Actions for Bipedal Locomotion
Просмотров 1,2 тыс.3 года назад
This is the video accompanying our ICRA 2021 submission: "Learning Task Space Actions for Bipedal Locomotion" Helei Duan, Jeremy Dao, Kevin Green, Taylor Apgar, Alan Fern, Jonathan Hurst arxiv.org/abs/2011.04741 Abstract- Recent work has demonstrated the success of reinforcement learning (RL) for training bipedal locomotion policies for real robots. This prior work, however, has focused on lear...
Sim-to-Real Learning of All Common Bipedal Gaits
Просмотров 3,9 тыс.3 года назад
Sim-to-Real Learning of All Common Bipedal Gaits
Cassie robot learns to hop, run and skip
Просмотров 16 тыс.3 года назад
Cassie robot learns to hop, run and skip
Learning Spring Mass Locomotion on Cassie
Просмотров 1,8 тыс.3 года назад
Learning Spring Mass Locomotion on Cassie
Learning Memory-Based Control for Bipedal Locomotion
Просмотров 1,3 тыс.4 года назад
Learning Memory-Based Control for Bipedal Locomotion
Planning for the Unexpected (ICRA 2020 Presentation)
Просмотров 2,8 тыс.4 года назад
Planning for the Unexpected (ICRA 2020 Presentation)
ATRIAS: An Agile and Efficient Bipedal Robot
Просмотров 14 тыс.9 лет назад
ATRIAS: An Agile and Efficient Bipedal Robot
ATRIAS Robot: 9.1 kph Running Speed (5.7 mph)
Просмотров 4,9 тыс.9 лет назад
ATRIAS Robot: 9.1 kph Running Speed (5.7 mph)
ATRIAS Robot: Traverses Rough Terrain at 6.6 kph (4.1 mph)
Просмотров 1,2 тыс.9 лет назад
ATRIAS Robot: Traverses Rough Terrain at 6.6 kph (4.1 mph)
ATRIAS Robot: Climbs a 15-cm Obstacle
Просмотров 19 тыс.9 лет назад
ATRIAS Robot: Climbs a 15-cm Obstacle
ATRIAS Bipedal Robot: Takes a Walk in the Park
Просмотров 171 тыс.9 лет назад
ATRIAS Bipedal Robot: Takes a Walk in the Park
ATRIAS Robot: Fail Friday - "Checked"
Просмотров 1,1 тыс.9 лет назад
ATRIAS Robot: Fail Friday - "Checked"
Great work! This is better than any college lecture. The fact you also make it work on hardware makes it admirable.
Thank you! This is a great presentation.
Does anyone know how to detect and visualize the points of the elevation map in the mujoco simulation in this video?
I think the robot may work efficiently by adding sensor at legs for obstacle detection
how does it respond when slightly pulled to left/right?
How to obtain the position of the box?
Cool
Я вижу прогресс: нет бессмысленного топтания на месте.
Much better. Love it
it's that auto, or by joystick?
The RL policy receives XY and yaw velocity commands from the user’s joystick to control the robot's heading and speed, while the policy manages the rest of the robot's responses.
I’m a lash tech in the UK but have a weird fascination with AI and was talking to Google Gemini about robotics and AI. Gemini recommended one of your videos to me to learn more about sim to real learning and I just wanted to say, human to human, this is so cool! It’s insane to see the progress you have made in the last few years and it’s so exciting to think of how we can use these robots one day for real world applications in which we can help humanity. Thank you for your amazing work ❤️
I am missing the side-by-side when the vision system is turned off, to see the qualitative improvement in walking behaviour.
This video document testifies that the machine, that all machines have the level of intelligence of a chicken.
@@danielsantini1153 From a motricity point of view yes, chicken still wins by a mile. From a cognitive point of view all robots have access to "ChatGPT/Gemini/LLama" level capabilities, which make them more knowledgeable and able to communicate with humans than any chicken I have seen. Yann Lecun has mentioned multiple times that "the dream" is to have machines able to "learn to be like a cat". This work is on that line of research. Looking are results like this ruclips.net/video/REvNnUzVDAA/видео.htmlsi=EdPmU4VdeWGnZV-u we seem to be on a good trajectory.
Thank you for your interest! You can refer to the papers here for examples of flat ground walking ruclips.net/video/AISy0hxo6-0/видео.htmlfeature=shared and stair walking ruclips.net/video/MPhEmC6b6XU/видео.htmlfeature=shared. Without visual inputs, the gaits either adapt to flat ground or maintain a higher swing height that always anticipates upcoming stairs. These examples should provide insight into the qualitative improvements in walking behavior when the vision system is turned off.
Interesting stuff, thank you for sharing it.
Only the beginning! This is like watching one of the first automobiles or airplanes. Great work by these very smart young scientists who will someday be creating bipedal robots that will be faster than human beings or even Ostriches (+43.5 miles per hour).
The Metal Gear is real
Full Speed In the Dark. It begins.
I don't know whats more frightening: This THING, or the comments section where nobody has considered how this can and will be weaponized against humanity. We don't need this shit. This has no use except to be running down humans. What the hell do you think will happen when these things become self aware and sentient, and decide we're a threat to them? This is horrifying.
CGI
PЯӨMӨƧM
Pog
Robots are cool who agrees
What software are y'all using?
From ATRIAS 1.0 to Cassie, OSU DRL has come a long way in making legged robots a reality.
this looks cgi its almost crazy
Very nice work!
Awesome! This is really impressive.
Cute
Vay vay vay
This is where the Corpus created their first ever weaponized M.O.A - Profits for Everyone!
In the future, I will see him pulling a plow and founding like the ancients, an empire with the strength of the wrist.
Reminds me of a toddler :)
I don't see any lasers, what kind of a show are we running here?
Now make a horse body then add legs to front and back
7랩 까지 계속 빨라지네 ㄷㄷㄷ
Some people keep saying, "I don't think that's a run.", but keep in mind the first robots to walk would make very small steps and took so long to move it was like a slideshow
This video is the starting point of all super cool things which could happen in the future.
I dont think thats running. Seems more like fast walking to me.
It seems a bit mean to expect Cassie to go up/down stairs (especially at that speed) with no way of knowing what to anticipate! (You try it while blindfolded!)
കുറച്ച് manushathyam kanikede pavatine thera para nadathikathe😢🙏
Evide chennalum kanumelloda oru malayali
0:15 - do robots need to go to bathroom?
Какая отвратительная штука
😂🤣🤦
It's running occasionally. Walking most of the time. Running both legs are up from ground.
very inspiring
Great job
rubber feet and put up red ones that it should not run over. that could be a child in the way. great.
Apply some rubbers on foot. It sounds like my grandma's sewing machine.
_This is frickin' A.M.A.Z.I.N.G._ I mean, look how natural its gait is! Replicating bipedal locomotion is an _incredible_ feat of engineering, and my hat's off to the monumental *geniuses* who accomplished this leap forward in robotics. 👏
they have cut a robo-gog in half and called it Cassie
Can’t be the only one creeped out by that thing
You aren't. Its fucking frightening. What nobody appears to get is this can be weaponized against humanity and WILL BE. You think for one second the powers in charge of pushing the development of this crap have our best interests at heart? I'm not that naive. This shit needs to be outlawed and they need to stop the production of these abominations.
Looks like Robotech BattlePod...