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 built a small biped robot you can teach new tricks to. Train it in simulation, run it on the real thing. Meet Microduck 🦆 $399, shipping before Christmas.

Pollen Robotics

19,784 subscribers

1,640,578 次观看 • 1 个月前 •via X (Twitter)

艺术科学技术教育

Anya Rossi• Live Now

Private livecam show

34 条评论

Evan Otero 的头像

Evan Otero1 个月前

These are the cutest things I’ve ever seen. Need about 100 running around the office.

Christian Bolivar 的头像

Christian Bolivar1 个月前

I need this thing following me around the dealership while I sell cars.

Norberto Marques 的头像

Norberto Marques1 个月前

Does the robot come with built-in survival training against toddlers? Otherwise I probably can’t have one around the house 😅

Li Yibai 的头像

Li Yibai1 个月前

I don’t need one but I absolutely need one 😭

Binh 的头像

Binh1 个月前

mannnnnnnnnn this is awesome i love it so much what do u mean it can even skate

Jeremy Nguyen ✍🏼 🚢 的头像

Jeremy Nguyen ✍🏼 🚢1 个月前

is there any way for me to get this in Australia @pollenrobotics ? (below: ships to Austria, but I don't see Australia anywhere)

でべ的头像

でべ1 个月前

I want to buy it from Japan too!

Aurora Zhang 的头像

Aurora Zhang1 个月前

"Grab" and "Roller Skater" tricks are so slept on. Deserve more orders if roller skater trick is included in launch

Elisym Labs 的头像

Elisym Labs1 个月前

It's 25cm tall with 15 motors and a camera. Ships before Christmas 2026, comes in 4 colorways. Comes with spare parts, NFC tags, a Hugging Face credit, and accessories like a laser pointer and rollers. Programmable in Python (and soon JavaScript, Scratch).

かきもち / kakimochi 的头像

かきもち / kakimochi1 个月前

日本への配送をサポートする予定はありますか？

SourceCodeplz 的头像

SourceCodeplz1 个月前

give it a wireless charging station, so it can charge by itself. and roam around non-stop.

Praveen Kumar Verma 的头像

Praveen Kumar Verma1 个月前

Is it modular? So that parts can be replaced if damaged?

JASON MCNAB 的头像

JASON MCNAB1 个月前

The huge miss. A self charging dock. Even vacuum cleansers self dock and charge.

RoboIndex 的头像

RoboIndex1 个月前

Congrats! Open-source robotics keeps getting more capable and accessible.

Gregor 的头像

Gregor1 个月前

the $399 gets you in but the real bet is whether the sim is faithful enough that behaviors actually transfer without retuning. that's where most hobby bots fall apart before you even touch contact physics.

Guru Medasani 的头像

Guru Medasani1 个月前

Ordered it just now. My kids love Reachy so excited for this one so they can play and train over Christmas holidays.

Evie Xu 的头像

Evie Xu1 个月前

ahhhhh adorable!!!

Diego Carlino 的头像

Diego Carlino1 个月前

I need this.

Marc The Shark 的头像

Marc The Shark1 个月前

What the what!?

Michael Waitze 的头像

Michael Waitze1 个月前

This is game-changing for robotics accessibility. The sim2real training pipeline you've built is exactly what the space needed. We actually broke down the full picture here:

Siddhartha Saxena 的头像

Siddhartha Saxena1 个月前

so cool!!!

฿ⱠɄeNine 9️⃣ 的头像

฿ⱠɄeNine 9️⃣1 个月前

Oh my god, I adore him. Phenomenal work!

Ezel Aras 的头像

Ezel Aras1 个月前

i don’t need one but unfortunately it has tiny shoes

Capitan Magnesio 的头像

Capitan Magnesio1 个月前

Lo que se encuentra Microduck la primera noche después de Navidad:

Martí Carmona 的头像

Martí Carmona1 个月前

banger

Broadstreet 的头像

Broadstreet1 个月前

Yes please, a great gift!

Nikhil Shahane 的头像

Nikhil Shahane1 个月前

Oh wow! This is really cool. My biggest concern in such cases is expectation vs reality!

Srini 的头像

Srini1 个月前

excited to see people training it to do stupid things and breaking it in public

฿ⱠɄeNine 9️⃣ 的头像

฿ⱠɄeNine 9️⃣1 个月前

Just sent your sales support team a question about a large duties fee that got tacked on during checkout to ship to California. 🖤

Suraj Jha 的头像

Suraj Jha1 个月前

Amazing !

Adam McFarlin 的头像

Adam McFarlin1 个月前

Love this!

OpenlabX 的头像

OpenlabX1 个月前

Complete details here

Lee Hayward 的头像

Lee Hayward1 个月前

I remember teaching people to inline skate at Hyde Park in London, back in the days. This little guy skating reminds me of the kids that would come for lessons. Made my day. Good luck with the launch!

Eidolon Research 的头像

Eidolon Research1 个月前

@oppen_9783 ça peut être cool pour tes enfants

相关视频

Train in sim. Run it for real. Teach it new tricks. 🦆 This is the sim2real loop that powers Microduck. RL stack open source here : Buy it here : Git : Join our community :

Train in sim. Run it for real. Teach it new tricks. 🦆 This is the sim2real loop that powers Microduck. RL stack open source here : Buy it here : Git : Join our community :

Pollen Robotics

326,836 次观看 • 1 个月前

waiting for the microduck robot to ship? 🦆 play with it on the microduck simulator space, running the same RL policies as the real robot ▶️

waiting for the microduck robot to ship? 🦆 play with it on the microduck simulator space, running the same RL policies as the real robot ▶️

Hugging Apps

28,486 次观看 • 1 个月前

The coolest part of Microduck might not be the robot. It's how it learns to move. This repo shows the full reinforcement learning pipeline behind Microduck. The robot learns in simulation using MuJoCo, PPO and GPU-based training, then the trained policy is exported to ONNX and deployed on the real robot. But the real challenge is sim-to-real. The simulation models things like actuator behavior, battery voltage, command delays, friction and even gear backlash. That helps make the learned policy work when it moves from simulation to the physical robot. The repo includes policies for walking, fall recovery, sitting, standing, kicking, ground picking and even roller skating. Train in simulation. Test the behavior. Export the policy. Run it on a real robot.

The coolest part of Microduck might not be the robot. It's how it learns to move. This repo shows the full reinforcement learning pipeline behind Microduck. The robot learns in simulation using MuJoCo, PPO and GPU-based training, then the trained policy is exported to ONNX and deployed on the real robot. But the real challenge is sim-to-real. The simulation models things like actuator behavior, battery voltage, command delays, friction and even gear backlash. That helps make the learned policy work when it moves from simulation to the physical robot. The repo includes policies for walking, fall recovery, sitting, standing, kicking, ground picking and even roller skating. Train in simulation. Test the behavior. Export the policy. Run it on a real robot.

Praveen Kumar Verma

138,017 次观看 • 1 个月前

Don't let the tiny duck fool you… there's a serious robot hiding inside. Meet Microduck from Pollen Robotics. It is a compact biped robot packed with 15 motors, a camera, ToF LiDAR, IMUs, microphones, a speaker and an articulated beak. And the best part? The software is open source. You can create behaviors, train policies in simulation and bring them to the real robot. A tiny duck learning to walk, crouch, grasp objects and recover from falls? That's not just cute. That's a pretty fun robotics platform.

Don't let the tiny duck fool you… there's a serious robot hiding inside. Meet Microduck from Pollen Robotics. It is a compact biped robot packed with 15 motors, a camera, ToF LiDAR, IMUs, microphones, a speaker and an articulated beak. And the best part? The software is open source. You can create behaviors, train policies in simulation and bring them to the real robot. A tiny duck learning to walk, crouch, grasp objects and recover from falls? That's not just cute. That's a pretty fun robotics platform.

Praveen Kumar Verma

358,737 次观看 • 1 个月前

Home robots are coming. The first ones in will be the small ones that are safe around kids and pets. Hugging Face's Pollen Robotics just opened pre-orders for Microduck, a $399 biped. 25cm, under 800g, 15 motors, a camera, a small LiDAR, two IMUs, and an articulated beak that picks things up. Ships with 7 pre-trained behaviors including walking, kicking, and roller skating. Open on GitHub under Apache-2.0: the SDK, the simulator, and the RL training pipeline. So you can teach it new tricks.

Home robots are coming. The first ones in will be the small ones that are safe around kids and pets. Hugging Face's Pollen Robotics just opened pre-orders for Microduck, a $399 biped. 25cm, under 800g, 15 motors, a camera, a small LiDAR, two IMUs, and an articulated beak that picks things up. Ships with 7 pre-trained behaviors including walking, kicking, and roller skating. Open on GitHub under Apache-2.0: the SDK, the simulator, and the RL training pipeline. So you can teach it new tricks.

The Humanoid Hub

779,029 次观看 • 1 个月前

come meet the microduck on oct 16th if you are in sf i may have convinced it to make its first ever appearance to the open public at the Open Together event 🦆

come meet the microduck on oct 16th if you are in sf i may have convinced it to make its first ever appearance to the open public at the Open Together event 🦆

apolinario (poli)

18,850 次观看 • 8 天前

We built a robot brain that nothing can stop. Shattered limbs? Jammed motors? If the bot can move, the Brain will move it— even if it’s an entirely new robot body. Meet the omni-bodied Skild Brain:

We built a robot brain that nothing can stop. Shattered limbs? Jammed motors? If the bot can move, the Brain will move it— even if it’s an entirely new robot body. Meet the omni-bodied Skild Brain:

Skild AI

2,445,111 次观看 • 1 年前

Look what our robot built!! Home StreetView FTW. All on the edge device -- if it can see it, it can clean it. If it can see it, it can manipulate it...(shh!!). WARNING: This is not a lab demo. It's a real robot, in a real home, solving real problems! Thanks to our incredible team Matic Robots, our fearless leader in Navneet Dalal, and incredible partner NVIDIA for enabling this!

Look what our robot built!! Home StreetView FTW. All on the edge device -- if it can see it, it can clean it. If it can see it, it can manipulate it...(shh!!). WARNING: This is not a lab demo. It's a real robot, in a real home, solving real problems! Thanks to our incredible team Matic Robots, our fearless leader in Navneet Dalal, and incredible partner NVIDIA for enabling this!

Mehul

59,347 次观看 • 2 年前

Asimov is now zero-shot sim2real for locomotion. You train the policy in simulation. It runs on the robot without additional tuning. Blog: Order yours:

Asimov is now zero-shot sim2real for locomotion. You train the policy in simulation. It runs on the robot without additional tuning. Blog: Order yours:

Asimov

28,243 次观看 • 1 个月前

Given the strong community adoption and real-world deployment of MolmoAct2 on YAM, we're introducing zero-shot evaluation of MolmoAct2 Bimanual YAM in simulation. Now you can test out our models without a real-world robot and build on them! Code: Simulation built on Maniskill!

Given the strong community adoption and real-world deployment of MolmoAct2 on YAM, we're introducing zero-shot evaluation of MolmoAct2 Bimanual YAM in simulation. Now you can test out our models without a real-world robot and build on them! Code: Simulation built on Maniskill!

Jiafei Duan

37,276 次观看 • 4 个月前

We gave a robot body to a GPT-4o AI agent. Here serving a glass from a human to another🍹 With only: • 30mn of arm demonstrations • A language prompt to guide it You can also teach it to clean, play, check in on you. What would you teach it? with Vignesh Anand

We gave a robot body to a GPT-4o AI agent. Here serving a glass from a human to another🍹 With only: • 30mn of arm demonstrations • A language prompt to guide it You can also teach it to clean, play, check in on you. What would you teach it? with Vignesh Anand

Axel

54,640 次观看 • 1 年前

i created microduck AR visualization so you can already see how it will look in your home and play a bit with it 🦆 link and instructions ⤵️

i created microduck AR visualization so you can already see how it will look in your home and play a bit with it 🦆 link and instructions ⤵️

apolinario (poli)

31,510 次观看 • 1 个月前

I placed 🥈 2nd in the LeHome Challenge (at IEEE ICRA 2026) earlier this month, and before that I was 🥇 1st of 62 teams in the simulation round. Now I am sharing my solution — with a detailed logic walkthrough and open-source code. The task was to teach a cheap two-armed robot to fold different garments in simulation and on a real robot. I trained a VLA policy with an RL loop to make it work. Let's break it down 👇

I placed 🥈 2nd in the LeHome Challenge (at IEEE ICRA 2026) earlier this month, and before that I was 🥇 1st of 62 teams in the simulation round. Now I am sharing my solution — with a detailed logic walkthrough and open-source code. The task was to teach a cheap two-armed robot to fold different garments in simulation and on a real robot. I trained a VLA policy with an RL loop to make it work. Let's break it down 👇

Ilia

29,522 次观看 • 3 个月前

15–30 minutes of real-world robot data. That's now enough to go from sim-to-real failure to working robot. Let’s see… You train a robot in simulation. You deploy it in the real world. It fails. The physics don't match. So you try to fine-tune it with real data, but… you never have enough real data, and the fine-tuning breaks everything the simulation taught it. SimDist fixes this with one key decision: don't transfer the policy. Transfer the world model. Keep the reward and value knowledge from simulation frozen. Only update the part that's actually wrong, how the robot predicts physics. Now the robot doesn't have to relearn the entire task in the real world. It already knows what success looks like. It just needs to correct its understanding of how the real world moves. The part that makes this work: they also trained on failures and recoveries; not just perfect demonstrations. Without that, the planner finds the gaps and exploits them. With it, the robot can tell a good future from a bad one. That's all it needs. Results on peg insertion, table leg assembly, locomotion on slippery and uneven surfaces. Tasks that require precision, force, and quick reaction. Thanks for sharing, Tyler Westenbroek ([Tyler Westenbroek]. Interactive visualization + paper: ——- Weekly robotics and AI insights. Subscribe free:

15–30 minutes of real-world robot data. That's now enough to go from sim-to-real failure to working robot. Let’s see… You train a robot in simulation. You deploy it in the real world. It fails. The physics don't match. So you try to fine-tune it with real data, but… you never have enough real data, and the fine-tuning breaks everything the simulation taught it. SimDist fixes this with one key decision: don't transfer the policy. Transfer the world model. Keep the reward and value knowledge from simulation frozen. Only update the part that's actually wrong, how the robot predicts physics. Now the robot doesn't have to relearn the entire task in the real world. It already knows what success looks like. It just needs to correct its understanding of how the real world moves. The part that makes this work: they also trained on failures and recoveries; not just perfect demonstrations. Without that, the planner finds the gaps and exploits them. With it, the robot can tell a good future from a bad one. That's all it needs. Results on peg insertion, table leg assembly, locomotion on slippery and uneven surfaces. Tasks that require precision, force, and quick reaction. Thanks for sharing, Tyler Westenbroek ([Tyler Westenbroek]. Interactive visualization + paper: ——- Weekly robotics and AI insights. Subscribe free:

Ilir Aliu

14,222 次观看 • 4 个月前

Physics simulation gives robots a place to learn before they enter the physical world. 🤖 Developers can model complex interactions, train and evaluate policies, and validate robot behavior in controlled virtual environments before deployment. Explore physics simulation for robotics 🔗

Physics simulation gives robots a place to learn before they enter the physical world. 🤖 Developers can model complex interactions, train and evaluate policies, and validate robot behavior in controlled virtual environments before deployment. Explore physics simulation for robotics 🔗

NVIDIA Robotics

28,742 次观看 • 20 天前

I’m still half in disbelief We asked Gemini Argon to teach the Apollo humanoid to play cornhole, and it zero-shotted it All we gave it was a video of a person doing it. It then took control of the robot, generated its own sim env, performed some sysID, and trained launches on the real robot. It took it a few real shots to improve and then land many consecutive airmails. The process was so long and articulated that we asked the robot itself to explain what it did to us, as you can see at the end of the video. It even came up with the gestures Great demo for a great model Google DeepMind Apptronik

I’m still half in disbelief We asked Gemini Argon to teach the Apollo humanoid to play cornhole, and it zero-shotted it All we gave it was a video of a person doing it. It then took control of the robot, generated its own sim env, performed some sysID, and trained launches on the real robot. It took it a few real shots to improve and then land many consecutive airmails. The process was so long and articulated that we asked the robot itself to explain what it did to us, as you can see at the end of the video. It even came up with the gestures Great demo for a great model Google DeepMind Apptronik

Norman Di Palo

216,236 次观看 • 1 天前

I finally have a Microduck again, so I was able to replay the emotions and compare simulation with the physical robot. This is early design, we can push this much further. What do you think? Music: Carefree, Kevin MacLeod

I finally have a Microduck again, so I was able to replay the emotions and compare simulation with the physical robot. This is early design, we can push this much further. What do you think? Music: Carefree, Kevin MacLeod

Remi Fabre

13,384 次观看 • 5 天前

🚀 We just shipped a big upgrade to our imitation-learning-in-simulation playground in LeRobot, built together with Lightwheel ! You can now teleoperate robots in sim (keyboard or real robot) and collect training demos instantly. This makes it possible to run real IL research on harder, more realistic manipulation tasks, even if you don’t have hardware. New tasks: 🟠 pick orange to the plate 🧺 fold cloth (yes!) 📦 pick 2 “e” toys to the box 🔴 lift red cube Through our partnership with Lightwheel, LeIsaac was integrated into EnvHub on day one, a best-practice integration that strengthens both ecosystems and pushes simulation-first robotics forward. 👉 Load a task via EnvHub, start teleop-ing, record your demo, upload the data on the hub and you’re ready to train or test.

🚀 We just shipped a big upgrade to our imitation-learning-in-simulation playground in LeRobot, built together with Lightwheel ! You can now teleoperate robots in sim (keyboard or real robot) and collect training demos instantly. This makes it possible to run real IL research on harder, more realistic manipulation tasks, even if you don’t have hardware. New tasks: 🟠 pick orange to the plate 🧺 fold cloth (yes!) 📦 pick 2 “e” toys to the box 🔴 lift red cube Through our partnership with Lightwheel, LeIsaac was integrated into EnvHub on day one, a best-practice integration that strengthens both ecosystems and pushes simulation-first robotics forward. 👉 Load a task via EnvHub, start teleop-ing, record your demo, upload the data on the hub and you’re ready to train or test.

LeRobot

44,896 次观看 • 10 个月前

Evaluating robot policies is hard. Ideally, instead of testing every new policy on a real robot, you could test in simulation; but simulations rarely correlate well with real-world performance. In order to make good, useful simulations, you need to spend a great deal of time and effort. That’s where PolaRiS comes in: it’s a toolkit that lets you take a short video of a real scene and turn it into a high-fidelity simulation. It provides what you need to build a good evaluation environment, and it “ships” with off-the-shelf environments that already show strong sim-to-real correlation, meaning that they can be used to inform policy performance. Arhan Jain and Karl Pertsch join us to talk about what they have built, why, and how you can use it. Watch Episode #62 of RoboPapers, with Chris Paxton and Jiafei Duan, now!

Evaluating robot policies is hard. Ideally, instead of testing every new policy on a real robot, you could test in simulation; but simulations rarely correlate well with real-world performance. In order to make good, useful simulations, you need to spend a great deal of time and effort. That’s where PolaRiS comes in: it’s a toolkit that lets you take a short video of a real scene and turn it into a high-fidelity simulation. It provides what you need to build a good evaluation environment, and it “ships” with off-the-shelf environments that already show strong sim-to-real correlation, meaning that they can be used to inform policy performance. Arhan Jain and Karl Pertsch join us to talk about what they have built, why, and how you can use it. Watch Episode #62 of RoboPapers, with Chris Paxton and Jiafei Duan, now!

RoboPapers

14,818 次观看 • 8 个月前

POV: You didn’t use Antioch to run test simulations on your robot before deploying it

POV: You didn’t use Antioch to run test simulations on your robot before deploying it

Historic Vids

346,539 次观看 • 4 个月前