Table tennis robot defeats some of world’s best players – why this has major implications for robotics

2026-04-28 · Source: Artificial intelligence (AI) – The Conversation · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Intermediate, long

Summary

Sony AI has developed Ace, a table tennis robot that has outperformed elite human players, marking a significant advance in AI systems operating in fast, uncertain, real-world environments. Ace won three out of five matches against competitive athletes with over ten years of experience, though it lost to professional Japanese league players. Unlike previous systems that simplified the game, Ace plays with standard equipment on a regulation table against human opponents using a full range of shots. Its performance is attributed to three innovations: event-based vision sensors and nine high-speed cameras for real-time ball tracking and spin estimation (up to 9,000 rpm), deep reinforcement learning trained in millions of virtual rallies for instant decision-making, and a high-performance robotic arm with two prismatic and six revolute joints for rapid, precise movements. This system demonstrates a meaningful reduction in the "sim-to-real" gap, enabling reliable operation under real-world uncertainty.

Key takeaway

For Computer Vision Engineers developing robotics for unstructured environments, Ace's success highlights the importance of integrating advanced perception (event-based vision), real-time decision-making (deep reinforcement learning), and robust physical actuation. You should explore combining learned control with optimization-based safety systems to bridge the sim-to-real gap, enabling robots to operate reliably and safely alongside humans in dynamic settings like manufacturing or healthcare.

Key insights

Ace robot demonstrates AI's ability to achieve superhuman performance in complex, real-world physical environments.

Principles

• Combine event-based vision with high-speed cameras for real-time tracking.
• Deep reinforcement learning enables instant decision-making in dynamic scenarios.
• Optimization-based control ensures safety and collision avoidance.

Method

Ace uses event-based vision and high-speed cameras for perception, deep reinforcement learning for decision-making trained in simulation, and a custom high-performance robotic arm for rapid, precise actions, all integrated with optimization-based control.

In practice

• Utilize event-based sensors for high-speed object tracking.
• Employ sim-to-real transfer for complex robotic tasks.
• Integrate learned components with safety-guaranteed control systems.

Topics

• Ace Robot
• Deep Reinforcement Learning
• Event-based Vision
• Sim-to-Real Transfer
• Human-Robot Interaction

Best for: Computer Vision Engineer, AI Scientist, Robotics Engineer, Research Scientist

Related on AIssential

• AI-powered robot beats elite table tennis players — Sony AI's Ace robot demonstrates advanced real-time perception and control in complex human-robot interaction.
• Meet Ace, the table-tennis robot that can beat elite players — High-speed perception systems enable robots to achieve elite performance in complex, dynamic tasks like table tennis.
• Game, Set, Bot: Sony AI’s ‘Ace’ Serves Up a Defeat to Table Tennis Pros — Gaming environments serve as safe, scalable testbeds for developing and benchmarking advanced physical AI agents.
• This robot can beat you at table tennis — The "Ace" robot demonstrates AI's ability to surpass elite human performance in complex physical sports.
• Benchmarking Action Spaces in Reinforcement Learning for Vision-based Robotic Manipulation — Joint velocity action space significantly improves sim-to-real performance for vision-based robotic picking and pushing tasks.
• Learning of Robot Safety Policies via Adversarial Synthetic Scenarios — An adversarial gamification framework enables robust robot safety policy learning by iteratively discovering high-risk edge cases.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial intelligence (AI) – The Conversation.