RAI Institute UMV Zero-Shot Deployment
Overview
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.
From RoboSpeak — WeChat
Event evidence and corrections
1 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
Environment: simulation
“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。”
Exact source · revision 2Source owner not reported
“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。”
Exact source · revision 1robospeak
“机器人自己在试错中摸索，怎么做能拿高分。”
Exact source · revision 1robospeak
“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。”
Exact source · revision 1robospeak
“所有运动策略在仿真环境中训练完成，直接部署到物理机器人，零样本迁移，无需现场调参、二次训练。”
Exact source · revision 2Source owner not reported
Reported quantity · speed: 8 other · Basis not reported · Differing source assertions
“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。”
Exact source · revision 2Source owner not reported
“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。”
Exact source · revision 2Source owner not reported
Reported quantity · height: 1 other · Basis not reported · Differing source assertions
“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。”
Exact source · revision 2Source owner not reported
“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。”
Exact source · revision 2Source owner not reported
Reported quantity · reliability: 15 other · Basis not reported · Differing source assertions
“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。”
Exact source · revision 2Source owner not reported
“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。”
Exact source · revision 2Source owner not reported
Reported quantity · weight: 23.5 other · Basis not reported · Differing source assertions
“整台机器人重23.5公斤，完全蹲伏时只有0.8米高。”
Exact source · revision 2Source owner not reported
“整台机器人重23.5公斤，完全蹲伏时只有0.8米高。”
Exact source · revision 2Source owner not reported
Reported quantity · height: 0.8 other · Basis not reported · Differing source assertions
“整台机器人重23.5公斤，完全蹲伏时只有0.8米高。”
Exact source · revision 2Source owner not reported
“整台机器人重23.5公斤，完全蹲伏时只有0.8米高。”
Exact source · revision 2Source owner not reported
Reported quantity · speed: 8 other · Basis not reported · Differing source assertions
“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。”
Exact source · revision 1robospeak
Reported quantity · height: 1 other · Basis not reported · Differing source assertions
“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。”
Exact source · revision 1robospeak
Reported quantity · reliability: 15 other · Basis not reported · Differing source assertions
“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。”
Exact source · revision 1robospeak
Reported quantity · weight: 23.5 other · Basis not reported · Differing source assertions
“整台机器人重23.5公斤，完全蹲伏时只有0.8米高。”
Exact source · revision 1robospeak
Reported quantity · height: 0.8 other · Basis not reported · Differing source assertions
“整台机器人重23.5公斤，完全蹲伏时只有0.8米高。”
Exact source · revision 1robospeak
Reported quantity · motor_count: 5 other · Basis not reported
“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。”
Exact source · revision 1robospeak
Reported quantity · motor count: 5 other · Basis not reported
“半秒钟，5个电机。”
Exact source · revision 2Source owner not reported
Evidence dependency changes (6)
- 2026-10-30T22:41:03.721Z · evidence updated · source revision 1. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T22:40:46.106Z · evidence updated · source revision 1. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T22:37:35.552Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T22:37:18.567Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T22:36:31.811Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T22:36:18.002Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
Report timeline
Follow attributed reports and material updates.
- RoboSpeak — WeChatSignalZero-Shot Real-World Deployment. RAI Institute Creates 5-Motor Bicycle Robot: 8 m/s, Half-Second Aerial Flips, 1-Meter Jump Platform
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.
- Chinese robotics — Research news discoveryZero-shot Deployment of a 5-Motor Bicycle Robot with Reinforcement Learning
Researchers from RAI Institute have developed a bicycle-shaped robot, UMV, capable of high-speed movement, jumps, and aerial flips using only five motors and reinforcement learning. The robot achieves 8 m/s speed, 1-meter jumps, and half-second aerial flips without specialized hardware for balance. Training occurs entirely in simulation, with zero-shot deployment to physical hardware. The study highlights the potential of reducing mechanical complexity and relying on learning algorithms for dynamic robotic tasks.
- Chinese robotics — Research news discoverySignalZero-Shot Real-World Deployment. RAI Institute Creates 5-Motor Bicycle Robot: 8 m/s, Half-Second Flips, 1-Meter Jump
The RAI Institute's UMV robot, a 23.5 kg bicycle-like robot with 5 motor-driven degrees of freedom, achieves high-speed movement (8 m/s), 1-meter jumps, and half-second aerial flips using reinforcement learning. All movement strategies are trained in simulation and directly deployed to the physical robot without additional tuning. The design philosophy focuses on reducing mechanical complexity and relying on algorithmic control for stability and dynamic capabilities.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).