Apple Research Team Unveils HALO Autonomous Control System for Humanoid Robots
苹果披露人形机器人进展 “大脑”已在宇树机器人上验证
On November 4, Apple's research team submitted a paper to arXiv detailing the HALO autonomous control system. Validated on Unitree G1 robots, the system transfers skills learned in simulation to unscanned real-world environments. The architecture utilizes a local visual language model for task planning and achieves an average success rate of 91% across single tasks in six different environments.
Source: Chinese robotics — Research news discovery · 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.
- Control
- Reported autonomousOpen source S2
Reported numbers
training_hours
≈48 hours
synthetic_data_hours
≈48 hours
test_trials
360
Reported trials
test_trials
10
Reported trials
GPU_hours
250
Reported duration
≈48 hours
GPU_units
64
GPU_units
4
- paper: Reported pending Open source S2
- code: not_reported
- weights: not_reported
- dataset: not_reported
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
研究团队将该系统部署在宇树科技的G1人形机器人上进行验证，让机器人将在仿真环境中学到的动作，带到陌生的现实房间中执行。
Open source S3
苹果披露人形机器人进展 “大脑”已在宇树机器人上验证 编辑于 2026-11-09 19:16 发布于： 北京市 11月4日，苹果研究团队在arXiv提交论文《Perceptive Humanoid Motion from Simulation》，公开人形机器人自主控制系统HALO。
Open source S2
据介绍，研究团队使用250张英伟达H100 GPU，耗时约48小时合成主要训练数据。
Open source S11
据论文披露，研究团队在6个未预先扫描或建图的真实环境中开展360次单项任务测试，平均成功率为91%。
Open source S8
另在一个真实环境中，对3种组合任务分别开展10次测试，平均成功率为70%。
Open source S9
动作生成器通过模仿学习训练，使用64张B200；动作跟踪器经强化学习、策略蒸馏等步骤训练，使用4张H100。
Open source S12
Source：Chinese robotics — Research news discovery · sohu.com