Outplaying elite table tennis players with an autonomous robot

Area of Science:

• Robotics
• Artificial Intelligence
• Human-Robot Interaction

Background:

• Artificial intelligence (AI) systems excel in computer games but struggle with real-world, high-speed physical interactions.
• Table tennis presents a significant challenge due to its demands for rapid, precise, and adversarial gameplay near physical limits.

Purpose of the Study:

• To develop the first real-world autonomous system, Ace, capable of competing with elite human table tennis players.
• To address the complexities of physical real-time interaction for AI agents.

Main Methods:

• Implemented a novel high-speed perception system utilizing event-based vision sensors.
• Developed a new control system employing model-free reinforcement learning.
• Integrated state-of-the-art, high-speed robotic hardware.

Main Results:

• Ace achieved competitive performance against elite and professional table tennis players under official rules.
• The system demonstrated consistent returns of high-speed, high-spin shots.
• Ace secured several victories in matches against human players.

Conclusions:

• Physical AI agents can successfully perform complex, real-time interactive tasks.
• Ace represents a significant advancement in autonomous systems for dynamic physical environments.
• The developed technologies hold potential for broader applications in human-robot interaction requiring speed and precision.