Research

Project Ace — robotic table tennis (Sony AI)

Contributor to a reinforcement-learning robotic table-tennis system that outplays elite human players, published in Nature.

Reinforcement learningPyTorchSimulationReal-time control

Concept visualization of the Project ACE robotic table-tennis system — Match footage preview
Dedicated space for approved Project ACE lab or match video

Concept close-up of Project ACE paddle hardware — Paddle assembly

Concept overhead view of Project ACE human evaluation — Human evaluation

Concept visualization of robotic paddle actuation — Actuation detail

Concept visualization of Project ACE tracking a ball — Perception loop

Problem

Table tennis demands sub-100ms perception, prediction, and control against a human opponent — a benchmark for real-time embodied reinforcement learning.

Approach

Reinforcement-learning policy for stroke selection and execution on a physical robot.
High-rate ball perception and trajectory prediction.
Sim-to-real transfer and extensive evaluation against human players.

Result

Published in Nature as "Outplaying elite table tennis players with an autonomous robot".
[EXACT PAPER TITLE AS PUBLISHED, YEAR, AUTHOR POSITION, DOI]
[YOUR SPECIFIC CONTRIBUTION — e.g. perception, simulation, hardware, evaluation]

System breakdown

Hardware

High-speed cameras
Robotic manipulator and paddle assembly
Table-tennis test rig

Firmware / embedded

Low-latency real-time control loop
Sensor synchronisation

Software

Reinforcement learning policy training
Simulation and sim-to-real pipeline
Evaluation harness

Project ACE — approved demo video slot

Note: Collaboration at Sony AI — exact role statement to be confirmed before publishing.

Discuss a similar system