Video yükleniyor...

Video Yüklenemedi

Ana Sayfaya Dön

Open Duck Mini is standing up using a policy learned with RL in simulation! We are still pushing locomotion, so if you have some knowledge of Isaac Gym and want to help feel free to join us! Thomas Wolf

antoinepirrone's profile picture

Antoine Pirrone

1,502 subscribers

66,082 görüntüleme • 1 yıl önce •via X (Twitter)

EğitimBilim & TeknolojiSanat
LIVE

Anya Rossi• Live Now

Private livecam show

11 Yorum

Antoine Pirrone profil fotoğrafı
Antoine Pirrone1 yıl önce

Actual link to join the LeRobot server : Look for the bdx-duck-robot-discussions channel :)

Loki Le Dev profil fotoğrafı
Loki Le Dev1 yıl önce

@Thom_Wolf Omg I've been wanting to make one for a while! I'm in vacation right now I'll come help when I'm back! Meanwhile if you have a proper urdf maybe @duburcqa could play with it in his simulator Looking forward to test on Isaac gym also.

Antoine Pirrone profil fotoğrafı
Antoine Pirrone1 yıl önce

@Thom_Wolf @duburcqa The urdf is here : :)

avi brown profil fotoğrafı
avi brown1 yıl önce

@Thom_Wolf instafollow

Appy Pie profil fotoğrafı
Appy Pie1 yıl önce

@Thom_Wolf Impressive work with Open Duck Mini and RL in simulation! 🚀 If anyone has experience with Isaac Gym and wants to contribute, this is a great opportunity. Excited to see how this advances locomotion capabilities!

Mark profil fotoğrafı
Mark1 yıl önce

@Thom_Wolf Cool 😎

Vai Viswanathan profil fotoğrafı
Vai Viswanathan1 yıl önce

@Thom_Wolf this duck is too cute

𝔭𝔬𝔢𝔩𝔩𝔩𝔩 profil fotoğrafı
𝔭𝔬𝔢𝔩𝔩𝔩𝔩1 yıl önce

@Thom_Wolf Love this!

Happy profil fotoğrafı
Happy1 yıl önce

@Thom_Wolf amazing bot

Hang Liu profil fotoğrafı
Hang Liu1 yıl önce

@Thom_Wolf Amazing！How do you overcome the sim2real gap when using digital servo? I guess digital servo only has position control but Isaac gym using torques to step simulation.

Antoine Pirrone profil fotoğrafı
Antoine Pirrone1 yıl önce

@Thom_Wolf That's a difficulty indeed. We model a simple PD controller in Isaac gym try to get the stiffness and damping values right, and add randomization during the training. Next step is to do a proper identification of the servos

Benzer Videolar

"Help us... If you want to cry, cry for those who have suffered because of the open border crisis..." Border Czar Thomas D. Homan responds to Selena Gomez's viral video on Trump's deportation policy. Watch below, and subscribe:
13:50

"Help us... If you want to cry, cry for those who have suffered because of the open border crisis..." Border Czar Thomas D. Homan responds to Selena Gomez's viral video on Trump's deportation policy. Watch below, and subscribe:

The Megyn Kelly Show

29,181 görüntüleme • 1 yıl önce

We are moving from model controller to RL for locomotion. The policy is starting to look quite robust. Even more convinced now that legs are easier than wheels, given the software complexity has shrunk tremendously with RL.
0:26

We are moving from model controller to RL for locomotion. The policy is starting to look quite robust. Even more convinced now that legs are easier than wheels, given the software complexity has shrunk tremendously with RL.

Sankaet

133,130 görüntüleme • 9 ay önce

Update on Open Duck Mini v2 ! The new version is more capable and much cheaper than the v1, and of course still fully open source ! We (with mansin and others) are now working hard on sim2real to achieve a nice walk using reinforcement learning, hopefully coming soon :)
0:19

Update on Open Duck Mini v2 ! The new version is more capable and much cheaper than the v1, and of course still fully open source ! We (with mansin and others) are now working hard on sim2real to achieve a nice walk using reinforcement learning, hopefully coming soon :)

Antoine Pirrone

27,354 görüntüleme • 1 yıl önce

We just open-sourced luwu_mjlab: an RL training environment for Luwu Dynamics quadruped robots. Built on mjlab + MuJoCo Warp, it lets us train locomotion and get-up policies in simulation, then validate them on real robots. GitHub:
0:32

We just open-sourced luwu_mjlab: an RL training environment for Luwu Dynamics quadruped robots. Built on mjlab + MuJoCo Warp, it lets us train locomotion and get-up policies in simulation, then validate them on real robots. GitHub:

XGO Robot

11,563 görüntüleme • 21 gün önce

🇧🇫 “Don’t send us food if you want to help us, send us tractors, fertilizer and materials so we can develop.” Thomas Sankara was assassinated
0:45

🇧🇫 “Don’t send us food if you want to help us, send us tractors, fertilizer and materials so we can develop.” Thomas Sankara was assassinated

Lord Bebo

844,589 görüntüleme • 1 yıl önce

Karl Popper on How We Gain Knowledge: ––Karl Popper: "People think, usually, that we acquire knowledge by opening our eyes and our ears and let the sensations stream into us, and they believe then that we record this like a camera. In my opinion, if we wish to get knowledge, we have to have a problem. It has to be knowledge of something. We have to find out something. We don't have to wait for information to stream into us, but we have to be inquisitive if we want to get knowledge. If we were passive, we would gain a confused mass of sensations or something like that, which we would hardly be able to understand and to convert into what one may call knowledge. Quite apart from that, perception is not really, in my opinion, the main source of our knowledge. The role of perception is to inform us about a momentary situation in our environment. But we couldn't really interpret our perceptions without knowing much more about our environment, namely, we know whether we are in a house or whether we are in a glacier. So we have two kinds of knowledge: this wider knowledge of a frame in which we orientate ourselves, and the momentary perception which gives us information about the situation at that particular moment. And it is only this situation in which we can use our perception. So we have theoretical knowledge and, if you like, the momentary practical challenge to our theoretical knowledge. And here comes perception in."
3:03

Karl Popper on How We Gain Knowledge: ––Karl Popper: "People think, usually, that we acquire knowledge by opening our eyes and our ears and let the sensations stream into us, and they believe then that we record this like a camera. In my opinion, if we wish to get knowledge, we have to have a problem. It has to be knowledge of something. We have to find out something. We don't have to wait for information to stream into us, but we have to be inquisitive if we want to get knowledge. If we were passive, we would gain a confused mass of sensations or something like that, which we would hardly be able to understand and to convert into what one may call knowledge. Quite apart from that, perception is not really, in my opinion, the main source of our knowledge. The role of perception is to inform us about a momentary situation in our environment. But we couldn't really interpret our perceptions without knowing much more about our environment, namely, we know whether we are in a house or whether we are in a glacier. So we have two kinds of knowledge: this wider knowledge of a frame in which we orientate ourselves, and the momentary perception which gives us information about the situation at that particular moment. And it is only this situation in which we can use our perception. So we have theoretical knowledge and, if you like, the momentary practical challenge to our theoretical knowledge. And here comes perception in."

Karl Popper Quotes

114,824 görüntüleme • 2 ay önce

There is a Black Ops REMASTERED mod currently in development for BO3. They are looking for more people to help out, so if you have BO3 modding knowledge and want to help, DM me and I can connect you to those behind the project.
1:15

There is a Black Ops REMASTERED mod currently in development for BO3. They are looking for more people to help out, so if you have BO3 modding knowledge and want to help, DM me and I can connect you to those behind the project.

Watchful Wolf

272,151 görüntüleme • 1 yıl önce

Have you noticed that something is happening to our country? The bullies are trying to take over. So, we’re standing up to them, and we need you with us. This Sunday, 7th September, march with us against antisemitism and extremism. Sign up now at
1:57

Have you noticed that something is happening to our country? The bullies are trying to take over. So, we’re standing up to them, and we need you with us. This Sunday, 7th September, march with us against antisemitism and extremism. Sign up now at

Campaign Against Antisemitism

28,013 görüntüleme • 10 ay önce

If you want to be apart of the largest grassroots political movement in American history, come join us in Kentucky. We are riding out for Thomas Massie May 14th to May 20th, to send a message to AIPAC and the Israel lobby. I pray you will come join us.
2:44

If you want to be apart of the largest grassroots political movement in American history, come join us in Kentucky. We are riding out for Thomas Massie May 14th to May 20th, to send a message to AIPAC and the Israel lobby. I pray you will come join us.

RyanMatta 🇺🇸 🦅

474,923 görüntüleme • 1 ay önce

Using our brain simulator, we’ve trained a reinforcement learning agent to maximize bits per second. Here is the RL policy converting brain data to cursor control in simulation:
0:29

Using our brain simulator, we’ve trained a reinforcement learning agent to maximize bits per second. Here is the RL policy converting brain data to cursor control in simulation:

Neuralink

18,410 görüntüleme • 1 yıl önce

Millions of people know Labour are failing us. If you want to join the revolt and help get rid of Labour, join us 🔗
0:34

Millions of people know Labour are failing us. If you want to join the revolt and help get rid of Labour, join us 🔗

Reform UK

87,006 görüntüleme • 1 yıl önce

Words, labels, repercussions, all irrelevant when our people and homeland are on the line. The preservation of our people is the highest law. “Your life doesn’t belong to you, it belongs to the race” - Thomas Sewell If you are tired of corruption and want to get active, we want you with us, join today.
2:17

Words, labels, repercussions, all irrelevant when our people and homeland are on the line. The preservation of our people is the highest law. “Your life doesn’t belong to you, it belongs to the race” - Thomas Sewell If you are tired of corruption and want to get active, we want you with us, join today.

WhiteVanguard

67,166 görüntüleme • 8 ay önce

🚨The home of Michael Jackson, Gary Indiana, there is nowhere we won’t go to bring some free stickers and cash, all we need for you to do is tell us where you want us to be next. We still have our foot on the gas.
0:28

🚨The home of Michael Jackson, Gary Indiana, there is nowhere we won’t go to bring some free stickers and cash, all we need for you to do is tell us where you want us to be next. We still have our foot on the gas.

Sticker Mule

47,289 görüntüleme • 2 ay önce

Unity doesn’t mean we’re going to join these radical leftists and sing kumbaya. We cannot unify with people who want us dead. Unity means reaching out to people who have woken up. If you recognize that the left is pushing this hatred and violence and refuse to be a part of it, then we welcome you to the family of faith, family, and freedom.
1:30

Unity doesn’t mean we’re going to join these radical leftists and sing kumbaya. We cannot unify with people who want us dead. Unity means reaching out to people who have woken up. If you recognize that the left is pushing this hatred and violence and refuse to be a part of it, then we welcome you to the family of faith, family, and freedom.

Kari Lake

66,489 görüntüleme • 9 ay önce

“They are pushing anti-White racism and socialism in South Africa. The worst of it is, we have a policy of Black Economic Empowerment. The reason Starlink isn’t in South Africa yet is because the South African government said to Elon Musk: if you want to bring Starlink to South Africa, you have to give 30% of your business to a Black person and we’ll tell you who.”
1:42

“They are pushing anti-White racism and socialism in South Africa. The worst of it is, we have a policy of Black Economic Empowerment. The reason Starlink isn’t in South Africa yet is because the South African government said to Elon Musk: if you want to bring Starlink to South Africa, you have to give 30% of your business to a Black person and we’ll tell you who.”

Taya Bass

22,594 görüntüleme • 2 ay önce

TRUMP IN ONE SINGLE EVENT -- "They won't be there for us" -- "We have some that are really enthusiastic. They're coming already." -- "This isn't 'need'" -- "If we need anything, they should be jumping to help us" -- "We want them to come and help us with the strait" -- My attitude is we don't need anybody"
0:42

TRUMP IN ONE SINGLE EVENT -- "They won't be there for us" -- "We have some that are really enthusiastic. They're coming already." -- "This isn't 'need'" -- "If we need anything, they should be jumping to help us" -- "We want them to come and help us with the strait" -- My attitude is we don't need anybody"

Aaron Rupar

940,365 görüntüleme • 3 ay önce

We are excited to share our #CORL2024 paper (oral) on "Learning Quadruped Locomotion Using Differentiable Simulation" done in collaboration with Sangbae Kim Massachusetts Institute of Technology (MIT). We present a new way to learn to walk in minutes without parallelization, outperforming PPO in sample efficiency! PDF: Video: We present a new framework for learning quadruped locomotion. By leveraging differentiable simulation for policy optimization, our approach achieves fast convergence and stable training, significantly outperforming model-free #ReinforcementLearning methods like PPO in sample efficiency. The key enabler is to combine a high-fidelity, non-differentiable simulator for forward dynamics with a simplified surrogate model for gradient backpropagation. Our framework enables learning quadruped walking in simulation in minutes without parallelization. When augmented with GPU parallelization, our approach allows the quadruped robot to master diverse locomotion skills on challenging terrains in minutes. This work highlights one of the first successful real-world applications of differentiable simulation for quadruped robots, offering a compelling alternative to traditional RL methods. Kudos to Yunlong Song! UZH Science University of Zurich UZH Space Hub UZH IfI European Research Council (ERC) Massachusetts Institute of Technology (MIT)MechE
1:13

We are excited to share our #CORL2024 paper (oral) on "Learning Quadruped Locomotion Using Differentiable Simulation" done in collaboration with Sangbae Kim Massachusetts Institute of Technology (MIT). We present a new way to learn to walk in minutes without parallelization, outperforming PPO in sample efficiency! PDF: Video: We present a new framework for learning quadruped locomotion. By leveraging differentiable simulation for policy optimization, our approach achieves fast convergence and stable training, significantly outperforming model-free #ReinforcementLearning methods like PPO in sample efficiency. The key enabler is to combine a high-fidelity, non-differentiable simulator for forward dynamics with a simplified surrogate model for gradient backpropagation. Our framework enables learning quadruped walking in simulation in minutes without parallelization. When augmented with GPU parallelization, our approach allows the quadruped robot to master diverse locomotion skills on challenging terrains in minutes. This work highlights one of the first successful real-world applications of differentiable simulation for quadruped robots, offering a compelling alternative to traditional RL methods. Kudos to Yunlong Song! UZH Science University of Zurich UZH Space Hub UZH IfI European Research Council (ERC) Massachusetts Institute of Technology (MIT)MechE

Davide Scaramuzza

15,533 görüntüleme • 1 yıl önce

Mobile-TeleVision follows our previous idea on upper-body and lower-body separation for control: 1⃣We train RL for lower-body locomotion and standing; 2⃣We use IK and retargeting directly for upper-body for high-precision and smooth manipulation. We avoid using RL for upper-body motion tracking. This is because fundamentally it is hard to avoid errors with RL tracking, and sim2real gap brings another gap. At this point, we still find using IK is more practical and precise.
1:15

Mobile-TeleVision follows our previous idea on upper-body and lower-body separation for control: 1⃣We train RL for lower-body locomotion and standing; 2⃣We use IK and retargeting directly for upper-body for high-precision and smooth manipulation. We avoid using RL for upper-body motion tracking. This is because fundamentally it is hard to avoid errors with RL tracking, and sim2real gap brings another gap. At this point, we still find using IK is more practical and precise.

Xiaolong Wang

32,662 görüntüleme • 1 yıl önce

🟢FIRST TRAILER IS UP🟢 if you are an artist (animator, vfx, background, etc) and want to help us out, feel free to send me a dm or fill out the form in the description. hope you like the trailer ü #sonicart #nazo #SonicTheHedegehog #animation
1:02

🟢FIRST TRAILER IS UP🟢 if you are an artist (animator, vfx, background, etc) and want to help us out, feel free to send me a dm or fill out the form in the description. hope you like the trailer ü #sonicart #nazo #SonicTheHedegehog #animation

🟢Team Full Power🟢

92,514 görüntüleme • 1 yıl önce