Robot plays ping pong with Ding Ning (2016 Olympic champion)
摘要
A robot played ping pong with 2016 Olympic champion Ding Ning, demonstrating its ability to generalize across different paddle orientations and positions.
This was cooler than the running events to me for two reasons: First, it alternates between forehand and backhand strokes. Second, when the paddle was placed in its hand, that precise orientation and position couldn't have been replicated exactly during training, since it's unlikely the robotic hand could grasp the paddle identically every time. This suggests the robot was trained to generalize to a range of possible orientations and positions—a crucial capability, because in ping pong, even tiny differences in the paddle's angle or position can determine whether the ball lands on the table or hits the ground.
相似文章
乒乓球机器人击败顶尖人类选手
一台乒乓球机器人已达到超人类水平,战胜了顶尖人类选手。
AGIBOT A3 现在在 BAAI 2026 大会上自主与人类进行乒乓球对战
AGIBOT A3 采用北京大学 SpikePingPong 算法和高频脉冲相机,在 BAAI 2026 大会上自主与人类进行乒乓球对战,展示了更快的视觉响应和高速动态控制能力,适用于工厂协作、服务机器人等实际应用场景。
玩乒乓球的神经形态算法
一种神经形态算法通过打乒乓球展示了学习能力和想象力,其表现优于标准算法,并能适应反转的控制指令。该代码使用 POWER-KI 和 PWK-AI-WORKBENCH 通过 100% 氛围编程构建。
@ChinaScience:看这个机器人在击败对手后调皮地跳舞。它不仅赢了,还跳起了胜利舞蹈……
一个机器人在击败对手后调皮地跳起了胜利舞蹈,展示了其有趣的行为。
我观看了一个机器人以真实速度跟上冰球,它预测比赛而非仅做出反应
本文讨论了机器人控制从反应式向基于预测的转变,重点介绍了LingBot-VA 2.0模型,该模型能够跟上快速移动的物体(如空气曲棍球),并且仅需少量演示即可学习。