Skild AI 利用 140 年的模拟自我比赛训练机器人踢足球
核心要点
- The robotics startup's S1 model learned to score goals by competing against itself in NVIDIA Isaac Sim, then transferred those skills to the real worl
The robotics startup's S1 model learned to score goals by competing against itself in NVIDIA Isaac Sim, then transferred those skills to the real world without any task-specific demonstrations.
Skild AI revealed on September 22 that its S1 model developed soccer-playing capabilities through a self-play training method conducted entirely within NVIDIA Isaac Sim. The training was focused on a single objective: score goals. No tailored rewards, no task-specific demonstrations, no hand-holding from human trainers. The resulting policy transferred directly to real-world scenarios, where the robot can now play football against both humans and other robots.
140 年是如何在短短几周内度过的
Modern GPU clusters can run thousands of parallel simulated environments simultaneously, compressing what would be over a century of real-time experience into a fraction of that in wall-clock time. Self-play, the technique Skild used, has a proven pedigree. DeepMind 曾用它来创建 AlphaGo 和 AlphaZero，这些系统通过与自身的副本竞争来掌握棋盘游戏。 The twist here is that Skild applied this approach not to a board game with discrete moves, but to the messy physics of a bipedal robot navigating a football pitch, involving continuous motor control, balance, object tracking, and real-time decision making.
格洛丽亚 看看内部人士和国会实际上在购买什么。凯莱财经揭示了值得重新审视的举措。追随金钱 →
机器人的训练目标被剥离到其本质。进球。 Everything else, the footwork, the positioning, the ball control, emerged as learned behaviors rather than programmed ones.
S1 车型和 Skild 更远大的抱负
Football is the flashy demo, but the S1 model’s real commercial value lies in its versatility. S1 于 2026 年 8 月下旬推出，可以通过单个视频演示执行长达 10 分钟的复杂操作任务。没有微调，没有参数更新。该公司展示了 S1 执行从煎饼翻转到套件组装等任务。
人工智能、科技及其所推动的市场——每日简报。日常的。自由的。加入加密货币、金融和政策领域 34,000 多名读者的行列。电子邮件地址 免费订阅 我们尊重您的隐私。随时取消订阅。
Skild AI’s development philosophy draws explicit parallels to large language models, training on extensive data from human videos and physics simulations to build general-purpose motor intelligence. NVIDIA 的模拟基础设施在此战略中发挥着关键作用。 Isaac Sim 提供物理引擎和渲染管道，使高保真机器人训练大规模可行，有助于缩小模拟与真实的差距。
140亿美元押注通用机器人
Skild reached a $100 million annual recurring revenue run rate shortly after its first commercial deployment earlier in 2026. Its robots currently operate across more than 60 client companies. Skild 在 2026 年 1 月的 C 轮融资中筹集了 14 亿美元，估值超过 140 亿美元。软银和 NVIDIA 参与了本轮融资。
ABB Robotics 和拥有 Universal Robots 和 MiR 的 Teradyne 都在与 Skild 合作。 Skild has also deployed robots assembling NVIDIA Blackwell GPU systems at Foxconn.