UC Berkeley recently released a video showcasing its Humanoid Table Tennis Robot, nicknamed HITTER, engaging in a high-speed game of table tennis with human opponents.

The robot demonstrated remarkable agility and precision, sustaining rallies of over 100 shots while maintaining balance with its left hand, emulating natural human movement.

HITTER’s performance relies on a two-part system: a high-level planner that tracks the ball and predicts trajectories, and a low-level calculator that translates those calculations into coordinated limb movements.

Trained on human motion data, the robot can pivot, swing, and move laterally with fluidity, reacting to balls traveling up to 5 meters per second in under a second, Interesting Engineering has reported.

Researchers combined model-based planning with reinforcement learning to enable HITTER to adapt to unpredictable shots, perform lifelike swings, and compete successfully against both humans and other humanoid robots in realistic match conditions.