Skip to content
RoboSpeak — WeChat· Dabaoo·· 2 hours agoSignalEditorial score91

Zero-Shot Real-World Deployment. RAI Institute Creates 5-Motor Bicycle Robot: 8 m/s, Half-Second Aerial Flips, 1-Meter Jump Platform

零样本真机直跑!RAI研究所造出5电机自行车机器人:8米/秒、半秒空翻、1米跳台

Summary

The RAI Institute's UMV robot, a bicycle-like machine with 5 motors, achieves 8 m/s speed, 1-meter jumps, and half-second aerial flips without specialized balancing hardware. All movement strategies are trained in simulation and directly deployed to the physical robot, demonstrating zero-shot transfer. The robot uses reinforcement learning to manage balance and dynamic maneuvers, with no springs, control moment gyroscopes, or dedicated stability mechanisms.

Editorial context

The paper presents a novel approach to robotic mobility by reducing mechanical complexity and relying on reinforcement learning for dynamic tasks. The UMV robot demonstrates zero-shot transfer from simulation to real-world deployment, achieving high-speed movement, jumps, and aerial maneuvers without specialized hardware for balance.

Source: RoboSpeak — WeChat · Read original article ↗

English translation unavailable. Choose the original article to read it.

What the source reports

Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.

Reported numbers

  • speed

    8

    View original evidence
    最高行驶速度: 8 m/s(约 28.8 km/h)
    Open source S21
  • height

    1

    View original evidence
    越障高度: 1 米
    Open source S21
  • reliability

    15

    View original evidence
    连续可靠性: 15 次自主跳台测试
    Open source S21
  • weight

    23.5

    View original evidence
    整台机器人重23.5公斤
    Open source S9
  • height

    0.8

    View original evidence
    完全蹲伏时只有0.8米高
    Open source S9
  • motor_count

    5

    View original evidence
    整台机器人只有5个驱动自由度
    Open source S6
Source excerpts and review record

Automatically extracted; no manual editorial approval recorded.

训练全程在NVIDIA Isaac Lab 仿真环境中完成 ,思路和教小孩学骑车很像,摔了就重来,做对了给奖励。

Open source S14

机器人自己在试错中摸索,怎么做能拿高分。

Open source S17

他用的方法叫迭代运动模仿(IMI),核心思路是,先生成一条物理上甚至不可行的粗略翻转轨迹当“草图”,让强化学习策略去模仿它,每迭代一次就优化一点,逐步收敛到真机可安全执行的动作。

Open source S22

最高行驶速度: 8 m/s(约 28.8 km/h),超过常规四足机器人的稳态巡航速度; 越障高度: 1 米,达到自身蹲伏高度的 130%; 连续可靠性: 15 次自主跳台测试,起跳阶段轨迹高度重合,波动只出现在落地冲击阶段。

Open source S21

整台机器人重23.5公斤,完全蹲伏时只有0.8米高。

Open source S9

一台23.5公斤的机器人在起跳瞬间的功率密度,接近一辆小型电动车的电机。

Open source S13

03. 七个月,几十次摔坏,换半秒空中翻转 论文里的实测数据,把这台极简机器人的能力拉到了非常直观的量级。

Open source S20

整台机器人只有5个驱动自由度,不到四足机器人(通常12个)的一半。

Open source S6

Source:RoboSpeak — WeChat · mp.weixin.qq.com

Timezone · UTC

Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).