Hugging Face 发布 399 美元开源双足机器人 Microduck
Hugging Face Unveils Microduck: A $399 Open-Source 25 cm Biped You Train with Reinforcement Learning
做机器人或 sim-to-real 的同学必看，399 美元就能拿到完整训练闭环，赶紧去 GitHub 抄 reward 和域随机化配方。
Most robotics launches ask you to trust a demo video. Pollen Robotics, the Bordeaux robotics team at Hugging Face, is instead shipping the training loop. This week it opened pre-orders for Microduck, a 25 cm bipedal robot in which every movement — walking, sitting, kicking, roller-skating, standing back up after a fall — is a neural policy trained in a physics simulator and exported to the hardware. It costs $399. The training environments, the reward functions, the domain-randomization settings, and the sim-to-real recipe are all public on GitHub. Microduck follows Reachy Mini, which has shipped more than 10,000 units, but reverses its premise: where Reachy Mini was built to sit on a desk and interact, Microduck is built to leave the desk, fall over, and get back up.
大多数机器人发布都要求你相信演示视频。相反，位于波尔多的Hugging Face机器人团队Pollen Robotics正在交付训练循环。本周，它开启了Microduck的预购，这是一款25厘米高的双足机器人，其每一个动作——行走、坐下、踢腿、轮滑、摔倒后重新站起——都是在物理模拟器中训练并导出到硬件的神经策略。售价399美元。训练环境、奖励函数、域随机化设置以及仿真到现实的配方都在GitHub上公开。Microduck紧随Reachy Mini之后，后者已出货超过10,000台，但Microduck颠覆了其前提：Reachy Mini被设计为放在桌上进行交互，而Microduck则被设计为离开桌面、摔倒并重新站起。
The Hardware
硬件
Microduck is 25 cm tall, 14 cm wide, and under 800 g. It carries 15 motors across legs, neck, and head, plus an articulated beak that picks objects off the floor. Compute is a Rockchip RK3566 with an AI accelerator, 1 GB of RAM, and 32 GB of storage.
Microduck高25厘米，宽14厘米，重量不到800克。它搭载了15个电机，分布在腿部、颈部和头部，外加一个可铰接的喙，用于从地面拾取物体。计算核心是带有AI加速器的Rockchip RK3566，配备1GB内存和32GB存储空间。
The sensor stack is unusually complete for the price. A front camera sits behind a dedicated camera-use indicator. Two IMUs are fitted, one in the body and one in the head. Range sensing is a compact LiDAR, an 8×8 time-of-flight matrix. There are microphones and a speaker, two NFC antennas, plus Wi-Fi and Bluetooth. Power is a removable NP-F550 battery, 2600 mAh, good for about an hour.
传感器配置在这个价位上异常完整。前置摄像头后面有专用的摄像头使用指示灯。配备了两个IMU，一个在机身，一个在头部。测距采用紧凑型LiDAR和8×8飞行时间矩阵。还有麦克风和扬声器、两个NFC天线，以及Wi-Fi和蓝牙。电源是可拆卸的NP-F550电池，容量2600mAh，可持续约一小时。
Seven trained moves ship in the box, driven by a bundled game controller before you write code: walk, sit and stand, kick, grab, roller-skate, and self-recovery. The robot does not speak. Each unit generates its own audio identity on first wake and keeps that voice permanently.
开箱即配有七种训练好的动作，在你编写代码之前，可通过附带的游戏控制器驱动：行走、坐下和站立、踢腿、抓取、轮滑和自恢复。机器人不会说话。每个单元在首次唤醒时生成自己的音频身份，并永久保留该声音。
How the behaviors are actually trained
行为实际上是如何训练的
Policies are trained in microduck_rl, built on mjlab (MuJoCo Warp) with PPO. Pollen reports roughly one to two hours on a CUDA GPU for a usable gait at 4096 parallel environments. Without a local GPU, appending --hf-jobs runs the same command on Hugging Face Jobs.
策略在microduck_rl中训练，基于mjlab（MuJoCo Warp）和PPO。Pollen报告称，在CUDA GPU上，使用4096个并行环境，训练出可用的步态大约需要一到两个小时。如果没有本地GPU，添加--hf-jobs参数可在Hugging Face Jobs上运行相同的命令。
The sim-to-real work sits in the actuator model. Each servo uses the BAM M6 model of the Dynamixel XL330 — voltage control law, back-EMF, and Coulomb, Stribeck, and load-dependent friction — rather than an ideal PD controller. Per-environment randomization covers battery voltage, voltage sag under load, command delay, and friction magnitude. Backlash variants train against ±1° of gear play, 2° total, in series with each of the 14 servo joints in the RL layout. Because the real encoder sits on the output side of that play, the observations read through it.
仿真到现实的关键在于执行器模型。每个舵机采用Dynamixel XL330的BAM M6模型——包括电压控制律、反电动势以及库仑、斯特里贝克和负载相关摩擦——而非理想的PD控制器。每个环境随机化涵盖电池电压、负载下的电压降、指令延迟和摩擦幅度。齿轮间隙变体在RL布局中每个14个舵机关节串联训练，对抗±1°的齿轮间隙，总计2°。由于真实编码器位于该间隙的输出侧，观测值通过它读取。
Trained policies export to ONNX with the observation normalizer baked into the graph. Pollen warns against deploying hand-converted checkpoints for exactly this reason.
训练好的策略导出为ONNX格式，并将观测归一化器嵌入图中。Pollen警告不要部署手工转换的检查点，正是因为这个原因。
On the robot, a Rust runtime drives the 50 Hz control loop and the motor bus. Every policy shares a 61-dimensional actor observation: 48 proprioception dimensions plus commands for twist (3), head pose (4), and body pose (6). That shared contract is what lets walk, recover, and trick policies hot-swap mid-run. Environments that ignore a command slot zero-pad it rather than dropping it.
在机器人上，Rust运行时驱动50Hz控制回路和电机总线。每个策略共享61维的actor观测：48维本体感觉加上指令（平移3维、头部姿态4维、身体姿态6维）。这个共享契约使得行走、恢复和特技策略可以在运行中热切换。忽略某个指令槽的环境会将其零填充，而不是丢弃。
The published registry covers 13 tasks: velocity tracking, stand-up, sit-stand, ground pick, ball kick (70 mm, 15 g ball, actor ball-blind), roulade, and five roller-skating environments.
发布的注册表涵盖13个任务：速度跟踪、站立、坐立、地面拾取、踢球（70毫米、15克球、actor盲球）、翻滚以及五个轮滑环境。
Key Takeaways
关键要点
- $399 open-source-software biped, pre-orders open August 27, 2026, deliveries targeted before Christmas.
- 15 motors, camera, LiDAR, two IMUs, NFC, Wi-Fi/Bluetooth, RK3566, ~1 hour runtime.
- Policies train in mjlab/MuJoCo Warp with PPO, ~1–2 hours for a gait at 4096 envs.
- Sim-to-real hinges on a BAM actuator model plus voltage, delay, friction, and ±1° backlash randomization.
- Software is Apache-2.0; the mechanical and electronic design files are not open.
- $399开源软件双足机器人，预购于2026年8月27日开放，交付目标在圣诞节前。
- 15个电机、摄像头、LiDAR、两个IMU、NFC、Wi-Fi/蓝牙、RK3566，续航约1小时。
- 策略在mjlab/MuJoCo Warp中使用PPO训练，4096个环境下的步态训练约需1-2小时。
- 仿真到现实的关键在于BAM执行器模型加上电压、延迟、摩擦和±1°齿轮间隙随机化。
- 软件采用Apache-2.0许可证；机械和电子设计文件未开源。
Check out the Microduck product page, launch blog post, press kit and spec sheet, microduck runtime repo, microduck_rl training repo and announcement from Thomas Wolf. Also, feel free to follow us on Twitter and don’t forget to join our 150k+ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well.
查看Microduck产品页面、发布博客文章、新闻资料包和规格表、microduck运行时仓库、microduck_rl训练仓库以及Thomas Wolf的公告。另外，欢迎在Twitter上关注我们，别忘了加入我们的150k+ML SubReddit并订阅我们的新闻通讯。等等！你在Telegram上吗？现在你也可以在Telegram上加入我们。
The post Hugging Face Unveils Microduck: A $399 Open-Source 25 cm Biped You Train with Reinforcement Learning appeared first on MarkTechPost.
文章《Hugging Face推出Microduck：一款$399开源25厘米双足机器人，用强化学习训练》首发于MarkTechPost。
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力