Skild AI Uses Self-Play to Train Robots to Master Physical Skills
Skild AI通过自我博弈训练机器人掌握物理技能
Skild AI has demonstrated a new self-play training method that enables robots to develop physical skills without relying on human demonstrations. The company used its S1 robot base model to play football in a simulated environment through over 140 years of adversarial learning. The study shows that robots naturally develop complex behaviors, such as walking, recovery, dribbling, and tackling, from a single objective like scoring a goal. The method is considered potentially capable of breaking through human capability limits and extending to everyday robotic tasks.
Skild AI's research highlights the potential of using self-play to train robots to master physical skills, with its method enabling robots to perform complex tasks like football in simulated environments without human demonstrations. The study emphasizes the transferability from simulation to reality and explores the possibilities of self-play in the field of physical AI.
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.
- Environment
- SimulationOpen source S3
Reported numbers
simulation_hours
>140
simulation_hours
>140
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
该公司利用其S1机器人基础模型，通过超过140年的模拟对抗学习踢足球。
Open source S3
经过超过140年的模拟对抗训练后，Skild将得到的策略从模拟环境迁移到实体人形机器人上，并展示了其踢足球的能力。
Open source S8
Source：Chinese robotics — Research news discovery · sohu.com