Robot plays tennis with humans in real time

A 4-foot humanoid robot powered by AI has been developed to play tennis with humans in real time, reacting instantly to shots without relying on pre-written scripts or remote control. Created by Chinese AI robotics company Galbot, this robot uses a system called LATENT, which trains it on imperfect human motion data. This approach enables the robot to learn movement fragments and combine them naturally, allowing it to execute coordinated, precise tennis shots and sustain rallies against human opponents. Unlike previous tennis-playing robots, this AI-driven humanoid demonstrates millisecond reaction times and full-body coordination, enabling competitive live play on both the forecourt and backcourt.

This development is significant because it showcases how robots can master complex athletic skills and human-like movements with incomplete data, advancing not only robotic tennis capabilities but also broader robotic interaction and adaptation in dynamic environments. The AI’s ability to observe, learn, and respond flexibly in real time highlights the progress in reinforcement learning and the application of simplified, approximate human motion for training robots. Demonstrations have included rallies where the robot consistently hits challenging shots, illustrating its growing proficiency and the potential for robotic teammates or competitors in sports and human activity contexts. This innovation marks a notable step forward in humanoid robot athletic performance and AI-driven physical interaction.