正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

We just released our work on robot soccer. I've been working on this for quite some time with my amazing colleagues at DeepMind. It's exciting how deep RL can produce such beautiful behaviors with low-cost robots. Full paper is available at Enjoy!

Tuomas Haarnoja

3,744 subscribers

758,752 次观看 • 3 年前 •via X (Twitter)

科学技术新闻政治教育

Anya Rossi• Live Now

Private livecam show

10 条评论

🌱Simon🌱 的头像

🌱Simon🌱3 年前

I never watch football, but I would pay a lot to watch a match between two robot teams.

Jouko Salonen 的头像

Jouko Salonen3 年前

Congrats! btw, I'm gonna do grandchild babysitting the whole coming 1st of May weekend. Do you happen to have a full 90 min video of this?

Nicholas Horbaczewski 的头像

Nicholas Horbaczewski3 年前

What, we’re just going to let that handball slide? The future of sports is robotic sports. Proud that @DroneRaceLeague is the first globally televised robotic sport.

Pieter Abbeel 的头像

Pieter Abbeel3 年前

very nice work @haarnoja, and fun to watch!

Karol Hausman 的头像

Karol Hausman3 年前

Some of these behaviors are beautiful! Congrats Tuomas!

Bernhard Jaeger 的头像

Bernhard Jaeger3 年前

Very cute robots. Feels much like small kids playing soccer.

Nolan 的头像

Nolan3 年前

I could dust both of these

Tony Hansmann 的头像

Tony Hansmann3 年前

Wow! How low-cost?

Erwin Coumans 🇺🇦 的头像

Erwin Coumans 🇺🇦3 年前

Very impressing on this low-cost hardware (Dynamixels?) @haarnoja , congrats!

Sélim 的头像

Sélim3 年前

so cute!! .. and obviously impressive 🤯 @FloMaushart

相关视频

Excited to share a new humanoid robot platform we’ve been working on. Berkeley Humanoid is a reliable and low-cost mid-scale research platform for learning-based control. We demonstrate the robot walking on various terrains and dynamic hopping with a simple RL controller.

Excited to share a new humanoid robot platform we’ve been working on. Berkeley Humanoid is a reliable and low-cost mid-scale research platform for learning-based control. We demonstrate the robot walking on various terrains and dynamic hopping with a simple RL controller.

Qiayuan Liao

80,294 次观看 • 1 年前

Happy to announce our neurips’25 paper, real world RL of active perception behaviors! I am pretty excited about this project - I learned that real world robot RL is actually quite straightforward. Details below:

Happy to announce our neurips’25 paper, real world RL of active perception behaviors! I am pretty excited about this project - I learned that real world robot RL is actually quite straightforward. Details below:

Edward Hu

24,565 次观看 • 7 个月前

🚀Meet Gemini Diffusion, our first diffusion-based and super fast language model, just announced at Google I/O!🚀 Very excited to be able to share what I've been working on for the past little while with our amazing small team Google DeepMind.

🚀Meet Gemini Diffusion, our first diffusion-based and super fast language model, just announced at Google I/O!🚀 Very excited to be able to share what I've been working on for the past little while with our amazing small team Google DeepMind.

Ivana Balazevic

43,472 次观看 • 1 年前

Working on my board for tonights game and every time I look up I see this face! How can I focus on work with this face looking at me?!?

Working on my board for tonights game and every time I look up I see this face! How can I focus on work with this face looking at me?!?

Kirk Herbstreit

883,705 次观看 • 1 年前

It's been a rollercoaster with some high highs and low lows, but I've loved my time getting to work and pour my passion into Lost Ark. 9yrs at Amazon, 8 at Amazon Games. 4 Games launched. See you on the next journey. o7

It's been a rollercoaster with some high highs and low lows, but I've loved my time getting to work and pour my passion into Lost Ark. 9yrs at Amazon, 8 at Amazon Games. 4 Games launched. See you on the next journey. o7

HardcoreHenry

55,279 次观看 • 3 个月前

Proud to share what I’ve been working on with my colleagues at Rhoda AI: Direct Video-Action Models (DVA). TL;DR: - We pre-train causal video models from scratch to control robots - They handle complex production tasks for hours without intervention - Only use ~10 hours of robot data How? 🧵👇

Proud to share what I’ve been working on with my colleagues at Rhoda AI: Direct Video-Action Models (DVA). TL;DR: - We pre-train causal video models from scratch to control robots - They handle complex production tasks for hours without intervention - Only use ~10 hours of robot data How? 🧵👇

Tongzhou Mu 🤖🦾🦿

19,102 次观看 • 3 个月前

Meet Cabinet: Paper Clip + KB. for quite some time I've been thinking how LLMs are missing the knowledge base - where I can dump CSVs, PDFs, and most important - inline web app. running on Claude Code with agents with heartbeats and jobs

Meet Cabinet: Paper Clip + KB. for quite some time I've been thinking how LLMs are missing the knowledge base - where I can dump CSVs, PDFs, and most important - inline web app. running on Claude Code with agents with heartbeats and jobs

Hila Shmuel

493,783 次观看 • 3 个月前

Learning Agile Soccer Skills for a Bipedal Robot with Deep Reinforcement Learning investigated the application of Deep Reinforcement Learning (Deep RL) for low-cost, miniature humanoid hardware in a dynamic environment, showing the method can synthesize sophisticated and safe movement skills making up complex behavioral strategies in a simplified one-versus-one (1v1) soccer game abs: project page:

Learning Agile Soccer Skills for a Bipedal Robot with Deep Reinforcement Learning investigated the application of Deep Reinforcement Learning (Deep RL) for low-cost, miniature humanoid hardware in a dynamic environment, showing the method can synthesize sophisticated and safe movement skills making up complex behavioral strategies in a simplified one-versus-one (1v1) soccer game abs: project page:

AK

293,335 次观看 • 3 年前

My weekend gift to you, my take on billie eilish’s “CHIHIRO” is now available for streaming on YouTube + SoundCloud only! I’ve been eager to share this with you for quite some time now 🙏🏿

My weekend gift to you, my take on billie eilish’s “CHIHIRO” is now available for streaming on YouTube + SoundCloud only! I’ve been eager to share this with you for quite some time now 🙏🏿

Black Coffee

62,762 次观看 • 1 年前

"A Bang at the Beach" a commission for: DaJester SFX pack: OpenNSFW A BIG commission I've been working on for quite some time now! I'm glad I can finally reveal it to you guys! 😚❤️❤️❤️ (✨HQ/download available through my bi0✨) #TheAmazingDigitalCircus #TADCEp7

"A Bang at the Beach" a commission for: DaJester SFX pack: OpenNSFW A BIG commission I've been working on for quite some time now! I'm glad I can finally reveal it to you guys! 😚❤️❤️❤️ (✨HQ/download available through my bi0✨) #TheAmazingDigitalCircus #TADCEp7

🔞✨Morado Arts✨(COMMS 3/5)

51,063 次观看 • 6 个月前

NEWS: DeepMind teaches robots to play soccer. In a paper released today Google's DeepMind details how they used Deep Reinforcement Learning (Deep RL) to train low-cost, miniature humanoid robots in dynamic environments, allowing them to play a simplified one-versus-one (1v1) soccer game. The robots, equipped with 20 actuated joints, were initially trained in simulation using the MuJoCo physics engine. Through this training, they learned robust and dynamic movement skills like rapid fall recovery, walking, turning, and kicking. The robots seamlessly transitioned between these skills, even surpassing expectations, and developed a basic strategic understanding of the game. During matches, the trained robots demonstrated agile skills such as turning, kicking moving balls, and dynamic defensive blocking. They quickly combined these skills, showcasing their adaptability and outperforming scripted baselines. The robots walked faster, got up quicker, and kicked faster than their counterparts. Individual skills were first trained in isolation within the simulation environment, and then composed in a self-play setting. The robots successfully transferred these skills to real-world scenarios. Full paper can be found here:

NEWS: DeepMind teaches robots to play soccer. In a paper released today Google's DeepMind details how they used Deep Reinforcement Learning (Deep RL) to train low-cost, miniature humanoid robots in dynamic environments, allowing them to play a simplified one-versus-one (1v1) soccer game. The robots, equipped with 20 actuated joints, were initially trained in simulation using the MuJoCo physics engine. Through this training, they learned robust and dynamic movement skills like rapid fall recovery, walking, turning, and kicking. The robots seamlessly transitioned between these skills, even surpassing expectations, and developed a basic strategic understanding of the game. During matches, the trained robots demonstrated agile skills such as turning, kicking moving balls, and dynamic defensive blocking. They quickly combined these skills, showcasing their adaptability and outperforming scripted baselines. The robots walked faster, got up quicker, and kicked faster than their counterparts. Individual skills were first trained in isolation within the simulation environment, and then composed in a self-play setting. The robots successfully transferred these skills to real-world scenarios. Full paper can be found here:

Dave Lee

372,039 次观看 • 3 年前

Q: "are you naturally calm or is it something you consciously work on?" Oscar: "both. it comes naturally for me, but it's also something i've been working on for a while. i'm not perfect, i'm not a robot, there are emotions there. it's just about how well you can control them."

Q: "are you naturally calm or is it something you consciously work on?" Oscar: "both. it comes naturally for me, but it's also something i've been working on for a while. i'm not perfect, i'm not a robot, there are emotions there. it's just about how well you can control them."

ari

26,327 次观看 • 1 年前

Everyone can now train high-dimensional muscle agents with RL, as we just released our code! Come join me at the #MyoChallenge workshop tomorrow at 10:00 in room 354 at #NeurIPS2023 to see my poster. Code is here!

Everyone can now train high-dimensional muscle agents with RL, as we just released our code! Come join me at the #MyoChallenge workshop tomorrow at 10:00 in room 354 at #NeurIPS2023 to see my poster. Code is here!

Pierre Schumacher

159,528 次观看 • 2 年前

Introducing 𝐀𝐋𝐎𝐇𝐀 𝐔𝐧𝐥𝐞𝐚𝐬𝐡𝐞𝐝 🌋 - Pushing the boundaries of dexterity with low-cost robots and AI. Google DeepMind Finally got to share some videos after a few months. Robots are fully autonomous filmed in one continuous shot. Enjoy!

Introducing 𝐀𝐋𝐎𝐇𝐀 𝐔𝐧𝐥𝐞𝐚𝐬𝐡𝐞𝐝 🌋 - Pushing the boundaries of dexterity with low-cost robots and AI. Google DeepMind Finally got to share some videos after a few months. Robots are fully autonomous filmed in one continuous shot. Enjoy!

Tony Zhao

353,928 次观看 • 2 年前

I've been working on procedural movement for my busy little duck and I'm stoked with how it's turning out! 🦆 #screenshotsaturday #gamedev

I've been working on procedural movement for my busy little duck and I'm stoked with how it's turning out! 🦆 #screenshotsaturday #gamedev

Mike Sebele

109,075 次观看 • 2 年前

I've been making movies with my best friends for 10 years. Now it's our full time job.

I've been making movies with my best friends for 10 years. Now it's our full time job.

Jason Carman

10,654 次观看 • 9 个月前

I've been working on deformable object manipulation since my PhD. It was totally a nightmare years ago and my PhD advisor was telling me not to work on it for my own good. Today, at ByteDance Seed, we are dropping GR-RL, a new VLA+RL system that manages long-horizon precise dexterous manipulation of deformable objects. This is probably the first real-world RL system to make a robot: ✅ Lace up your shoes end to end ✅ Hit millimeter tolerance repeatedly ✅ Recover from mistakes (See video!) ✅ And complete continuous shoelace threading on a real bimanual platform 📈 Success rate: ↑ from 45.7% → 83.3% Yes, robots can now actually do this. Project page: ArXiv:

I've been working on deformable object manipulation since my PhD. It was totally a nightmare years ago and my PhD advisor was telling me not to work on it for my own good. Today, at ByteDance Seed, we are dropping GR-RL, a new VLA+RL system that manages long-horizon precise dexterous manipulation of deformable objects. This is probably the first real-world RL system to make a robot: ✅ Lace up your shoes end to end ✅ Hit millimeter tolerance repeatedly ✅ Recover from mistakes (See video!) ✅ And complete continuous shoelace threading on a real bimanual platform 📈 Success rate: ↑ from 45.7% → 83.3% Yes, robots can now actually do this. Project page: ArXiv:

Xiao Ma

109,462 次观看 • 7 个月前

Rad New World☢️ [Dialogue System] This little project is something I've been tinkering with on and off for quite a while now — it's basically my love letter to some of my all-time favorite games: Fallout 1,2 & Tactics. Cheers!🫡

Rad New World☢️ [Dialogue System] This little project is something I've been tinkering with on and off for quite a while now — it's basically my love letter to some of my all-time favorite games: Fallout 1,2 & Tactics. Cheers!🫡

Frontline: Games Series

29,488 次观看 • 1 个月前

"RAD New World☢️" [Concept] This little project is something I've been tinkering with on and off for quite a while now — it's basically my love letter to some of my all-time favorite games: Fallout 1,2 & Tactics. Cheers!🫡

"RAD New World☢️" [Concept] This little project is something I've been tinkering with on and off for quite a while now — it's basically my love letter to some of my all-time favorite games: Fallout 1,2 & Tactics. Cheers!🫡

Frontline: Games Series

32,325 次观看 • 3 个月前

How can robots acquire fine-grained manipulation skills? Introducing ACT: Action Chunking with Transformers 🤖 Key idea: Imitation, but predict actions in chunks instead of one at a time. Here are results with only ~15min of demonstrations, running on low-cost arms:

How can robots acquire fine-grained manipulation skills? Introducing ACT: Action Chunking with Transformers 🤖 Key idea: Imitation, but predict actions in chunks instead of one at a time. Here are results with only ~15min of demonstrations, running on low-cost arms:

Tony Zhao

246,823 次观看 • 3 年前